You are on page 1of 11

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1, pp.

25-35 (2007)

LARGE EDDY SIMULATION OF THERMAL JETS IN CROSS FLOW


F. Ma * and, M. Satish, M. R. Islam
Civil and Resource Engineering Department, Dalhousie University
Halifax, Nova Scotia, Canada, B3H 4R2
*
E-mail: fma@dal.ca (Corresponding Author)
ABSTRACT: This study is related to the numerical simulations of a bent-over thermal jet. The governing equations of
the fluid are solved with the help of a three-dimensional buoyancy-extended large eddy simulation (LES) numerical
model. In addition, the dynamic procedure is used to evaluate the Smagorinsky model coefficient. The finite difference
formulations of the governing equations are split into three parts related to advection, dispersion and propagation. The
advection part is solved by the QUICKEST scheme. The dispersion part is solved by the central difference method and
the propagation part is solved implicitly by using the Gauss-Seidel iteration method. The initial turbulence of the
thermal jet from the orifice is accounted for by introducing random disturbances to the flowing parameters. Ensemble
averaged determinant relationships for the bent-over jet trajectory, jet sizes and concentration dilution are presented. The
salient characteristics of the bent-over thermal jet are captured, including variability among different realizations of the
thermal jet, the development of protuberances, the horseshoe cross sectional shape and the hollow trough along the bentover concave side of the jet. The protuberance characteristic and the asymmetric shape of the jet from the present study
are compared with the results from the conventional k- model. The horseshoe cross sectional shape and the trough or
bifurcation characteristics are investigated by studying the inner structures of the flow field. These quantitative
relationships and qualitative observations are found to be in good agreement with experimental results from an earlier
investigation.

1. INTRODUCTION
stagnant fluid and found that the motion of a
thermal resembles that of a vortex ring. Koh and
Chang [4] employed an integral model and
predicted the bulk flow characteristics of a thermal.
The behaviour of a thermal was numerically studied
using a two-dimensional k- model and later by Ma
and Li [5] using a three-dimensional k- model.
However, in their studies, only average bulk
characteristics of the thermal were reported, the
variability from one run to another and the
protuberance characteristics of the thermal were not
captured. The variability and protuberance
characteristics are salient features of a thermal
observed in both laboratory and engineering
practice. Recently, Li and Ma [6] employing the
large eddy simulation (LES) successfully
reproduced certain variability characteristics of an
instantaneous source thermal in stagnant fluid.
The behaviour of a thermal jet in the presence of a
cross flow is much more complicated than in a
uniform stagnant ambient environment. And, in
coastal regions, where a tidal current exists most of
the time, knowledge of the behaviour of a thermal
jet in an ambient current is an essential prerequisite
for developing efficient outfalls and also for
assessing the behaviour of existing outfalls. The

Wastewater in our society is inevitably returned to a


natural water body, usually in the form of sewage
outfalls. This makes use of the characteristics of a
buoyant or thermal jet resulting in an effective
mixing process within the ambient fluid. Despite
the wide application of buoyant jets and its great
impact on our environment, its behaviour is quite
complicated and not yet fully understood. A
thorough understanding of the process involves the
knowledge of the behaviour of two different and
interacting turbulent fluid flows. Great efforts have
been made in predicting the behaviour of
discharged wastewater, including the determination
of the pollutants transport, mixing and dilution. In
recent years, considerable experimental and
numerical efforts have been carried out and many
basic findings have been reported. For example,
Scorer [1] discovered that thermals move in an
enveloping cone of fixed angle in stagnant ambient
fluid. However, the angle of the enveloping cone
was found to differ significantly among different
thermals and even among different realizations of a
certain thermal. Turner [2] also found that there
exists significant variability among different
realizations of a thermal. Woodward [3]
investigated the velocity distribution of a thermal in
Received: 5 May 2006; Accepted: 6 Nov. 2006
25

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

objective of the present study is to investigate using


the LES approach, the basic mechanisms of a
thermal jet in cross flow including not only its
overall determinant relationships but also the
variability features.

Denoting time by t, hydrodynamic pressure by p ,


fluid density by ref and scalar density difference
by , the incompressible buoyancy extended
momentum equations are:
ui ui u j
2 ui ij
1 p

+
=
+
+
+ gi
t
x j
x j x j x j
ref xi
ref

2. NUMERICAL METHODOLOGY

(i = 1,2,3)

(2)
in which and g i are fluids molecular viscosity
and gravitational acceleration respectively. The last
term on the right hand side of equations (2)
accounts for the buoyancy effect of the fluid [8].

2.1 Governing equations


A typical definition sketch of a thermal jet in
flowing water is illustrated in Fig.1. The turbulent
flow of incompressible fluid is governed by the
Navier-Stokes equations. Due to the highly frequent
fluctuation, inherent structure of the turbulent flow,
great computational efforts are required to carry out
a direct numerical simulation (DNS). Also, the
conventional k- type of turbulence model, solving
the Reynolds-averaged Navier-Stokes (RANS)
equations, can only give average bulk solutions; it
cannot capture the variability features. To deal with
the variability characteristics of a thermal jet, the
buoyancy extended LES approach is adopted in this
study.

The sub-grid scale stress is ij = ui u j ui u j . By


using the eddy viscosity assumption, Boussinesqs
hypothesis, it can be parameterized as:

ij +
in

ij
3

kk = 2 t S ij

which

the

(3)
fluid

share

strain

is

. By using the mixing length

concept, the eddy viscosity t in equation (3) can


S ij =

1 u i u j

+
2 x j x i

be modeled as:

t = L2s S

where

the

by S =

With a spatial filter of width applied to the


incompressible Navier-Stokes equations, LES
approach has the following governing equations [7]:

by

ui

Ls = C s

strain

rate

is

defined

2 S ij S ij . In the LES approach, the

(5)

The proportional coefficient C s adopted by Li and


Ma [6] for equation (5) was assumed to be a
constant. In fact, this coefficient varies both in
space and time [10]. In the present study C s in
equation (5) is evaluated by the so called dynamic
model, the details of which are presented in the
following section.

and

coordinates by xi ( i = 1,2,3 ) in i direction, the


incompressible continuity equation has the
following form:

u i
=0
xi

local

modeling of mixing length Ls in equation (4) is


based on the Smagorinsky [9] model and assumed
to be proportional to the sub-grid scale
characteristic length :

Fig.1. Definition sketch of a buoyant jet

i) Continuity equation
Denoting velocity component

(4)

(1)

iii) Dynamic model


Following Germano et al. [11]), a second filter of

in which the Einsteins convention of summation


over repeated indices are adopted and the over bar
denotes the application of the spatial filter.

width , larger than the first filter width , is


applied to equations (1) and (2). For the sake of
brevity, the resulting equations will not be rewritten here, they are similar to equations (1) and

ii) Buoyancy extended momentum equations

26

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

in which St is Schmidt number and taken to be 0.5


[16]; and u, v and w are velocity components in x, y
and z directions respectively.
The buoyancy fluids density is related to its
concentration by

(2). In the double filtered momentum equation, the


sub-grid scale tensor of the field u that must be
modeled

is

Tij = u i u j u i u j

Assuming

Lij = u i u j u i u j to be the resolved turbulent

= ref +

stress corresponding to the second filter applied to


the field u , we have

Lij = Tij ij

1
.
in which =
ref C

(6)

Following the least square approach of Lilly [12],


the model coefficient C s in equation (5) is
expressed as

Cs =
with

1 Lij M ij
2 M kl M kl
2

= (xyz )

in

which

and x , y and z are grid sizes


in x, y and z directions respectively. Based on
Lesieur and Metais [10], an integer value of 2 is
3

chosen for the ratio of

(9)

2.2 Numerical algorithm


To seek numerical solutions to the above governing
equations (1)-(9), the split operator approach [7] is
used so that different physical processes can be
accounted for by suitable numerical schemes. At
each time increment, the equations are split into
three steps: advection, diffusion and propagation.
The advection part is solved by the QUICKEST
scheme [17], which is thirdorder accurate. The
dispersion part is solved by the central difference
method. And in the propagation part, the Poisson
type equation is solved implicitly by using the
Gauss-Seidel iteration method.

(7)

M ij = S S S S

( ref ) = ref 1 + ( ref )


C

in the present study.

The model coefficient C s given by expression (7) is,


in principle, a function of both spatial directions x, y,
z and time t. Unfortunately, analysis of both DNS
data [13] and experimental data [14] reveals that the
C s field predicted by this expression varies
strongly in space and contains a significant fraction
of negative values. The technique suggested by [15]
of removing the local indeterminacy attached to
expression (7) is adopted in the present study. The
numerator and the denominator of expression (7)
are averaged over the horizontal direction to
eliminate ill-conditioning associated with small and
negative values of the denominator.

2.3 Boundary conditions


At solid boundaries the slip velocity is estimated by
the log-law. For inlet and outlet boundaries, the
periodic boundary condition is used [18]. The
imposition of free surface boundary condition is
relatively difficult in LES. In the current work, this
problem is overcome by applying the model to an
equivalent problem of flow through a closed
conduit with vertical dimension equal to twice the
water depth.
Jet turbulence from the orifice is implemented by
superposing random numbers on both average bulk
velocity and scalar concentration fields [6]. In view
of our numerical experiments and the work down
by [19], a random number, of magnitude of 10% of
Uj, is introduced to the jet orifice velocity field,
where Uj is jets average bulk discharge velocity;
and a random number, of magnitude equal to 6% of
the average thermal concentration, is introduced to
the jet orifice concentration field.

iv) Scalar transport equation


The density variation in momentum equations (2) is
generally related to certain scalar quantities, such as
temperature. Denoting scalar concentration by C,
the filtered mass conservation of a scalar is given
by
C
C
C
C
+u
+v
+w
=
t
x
y
z

t C
t C
t C
+
+

+
+
+
x
St x y
St y z
St z

2.4 Initial conditions


The fully developed initial condition of the open
channel flow is achieved by running the program
for a predetermined period of time without the jet
intruding. To obtain the ensemble average results of

(8)

27

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

jet trajectory, jet sizes and concentration dilution


are presented in the following sections.

the thermal jet, a series of realizations of the


thermal jet are carried out. Numerical trials from
the present investigation indicated that generally a
series of five realizations were sufficient to perform
an ensemble averaging process. Additional
realizations made only a marginal difference to the
overall averaged results. The time interval for
consecutive realizations was taken as 5T to
eliminate the effect of the previous realization on

i) Jet trajectory
Fig.2 shows five different realizations of the jet
trajectory, and Fig.3 presents their ensemble
average result together with its corresponding
regression

X
Zd
= c1
Lmv
Lmv

L
the channel flow, where T =
and L is channel
U0

(10)

length. As the volume of jet is very small compared


to the volume of fluid in the open channel, its effect
on the global flow should be negligible.

where Zd is the downward jet displacement;


(QU j Sin )1 2
L mv =
is a length scale describing the

3. RESULTS AND DISCUSSIONS

bent-over of momentum of the jet in cross flow and

Q = D 2U j is volumetric discharge of the jet. And

U0

3.1 Simulation parameters


The channel dimensions used in the numerical
simulation are 0.8m long, 0.4m wide and 0.4m deep
in x, y and z directions respectively. It should be
noted that the length of the channel used for the
numerical simulation is shorter than the one used in
the physical experiment [20] to reduce
computational effort. It is expected that the effect of
shortening the channel length is not significant,
especially when the periodic boundary condition is
employed. For the purpose of comparing the present
numerical results with experimental results all the
other parameters chosen for the present study are
the same as those used by Wong [21] in his physical
experiment. The channel and jet velocities, U0
(Fig.1) and Uj, are 0.05m/s and 0.575m/s
respectively. Jet orifice diameter D is set at
0.00755m. A value of = 85 is selected to ensure
no momentum excess in x direction between the jet
and the ambient flow. The total number of grids
used in the computation is 16080160. The ratio

= 0.005% .
of the density difference is
ref
A series of five realizations were carried out, and
the results of the realizations both in terms of
quantitative and qualitative observations are
discussed and compared with [21] experimental
results in the following sections.

c1 and 1 in equation (10) are the coefficient and


the exponent respectively. Their regression results
were found to be c1 = 1.46 and 1 = 0.34 . From
dimensional analysis and experimental investigation
[21] obtained a value of 1 = 0.34 and average

Zd/m

value of coefficient c1 = 1.56 (Table 1).

Fig.2. Jet trajectories for five different realizations

Zd
Lmv

X
L mv

Fig.3. Ensemble average jet trajectory (dot) and its


regression (line)

ii) Jet sizes


Fig.4 shows jet sizes, vertical and horizontal
diameters of the cross section Dz and Dy. For
brevity, only two realizations are plotted. However,
the ensemble average results from the total of five
realizations are approximated as:
Dz = c2 Z d
(11)

3.2 Ensemble averaged determinant


relationships
The simulated quantitative results of a series of five
realizations are shown in Figs.2-5. Ensemble
averaged quantitative relationships for estimating
28

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

and

D y = c3 Z d

iii) Concentration dilution


Fig.5 gives dilution ensemble average result as
shown in equation (13)

(12)

where c 2 and c3 are coefficients. Regression results


from the present study for the two coefficients are:
c 2 = 0.53 and c 3 = 0.66 respectively. It is
interesting to note that the horizontal size is greater
than the vertical size in the bent-over thermal jet.
The average horizontal to vertical size ratio

X
SQ
= c 4
2
U 0 Lmv
Lmv
where S =

Dy

Dy
Dz

than that in the

downstream part. For the vertical size equation (11),


Wong [21] from his experiments reported the
coefficient c 2 to be between 0.49 and 0.62, with an
average value of 0.55. He reported no results for the
horizontal diameter given by equation (12). But he
obtained the ratio of

Dy

to be between 1.05 and

Dz

(13)

is ensemble average minimum

dilution and Cmax is maximum concentration in a


cross section of the thermal jet; and c 4 and 2 are
coefficient and exponent respectively. Regression
results for the coefficient and exponent are:
c 4 = 1.10 and 2 = 0.68 . Again, from dimensional
analysis and experimental data [21] concluded that
the exponential power 2 in equation (13) to be
0.66 and the coefficient c 4 to be between 0.88 and
1.31, with an average value of 1.13.
The ensemble average results for the thermal jet are
shown tabulated along with the experimental results
reported [21] in Table 1. The numerical results
from the present study show good agreement with
the reported experimental results related to jet
trajectory, jet sizes and concentration dilution.

Dz

was computed to be around 1.25. However, a


section located in the upstream part of the jet has a
slightly lower value of

1
C max

Dz/m

1.36, with an average value of 1.23.

Z/m

Z/m

(a) Diameter in Z-direction

(b) Diameter in Ydirection


Fig.4. Jet sizes for two different realizations

Table 1. Jet trajectory, horizontal and vertical sizes, and concentration dilution
Item
Jet
trajectory

Jet sizes

Concentration
dilution

Relationship

Experiment
(Wong, 1991)
average
range

Parameters

Computation

c1

1.46

1.56

1.41~1.81

0.34

1/3

fixed

Vertical D z = c 2 Z d

c2

0.53

0.55

0.49~0.62

Horizontal D y = c 3 Z d

c3

0.66

Size ratio D y / D z

c3 / c 2

1.25

1.23

1.05~1.36

c4

1.10

1.13

0.88~1.31

0.68

2/3

fixed

X
Zd
= c1
Lmv
Lmv

SQ
U 0 L2mv

X
= c 4
L mv

29

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

And these protuberances will affect the subsequent


motion and behaviour of the thermal, causing
asymmetric patterns of the thermal jet. The
numerical capability of capturing this feature is
largely due to the random disturbances introduced
to the orifice boundary condition as elaborated in
the boundary condition section.

SQ
U 0 L2mv

X
Lmv
Fig.5. Ensemble average minimum dilution (circle) and
its regression (line)

Z/mm

3.3 Salient characteristic features


i) Variability
From the process of deriving the above ensemble
averaged relationships, we know that different
realizations of the thermal jet have different jet
trajectories (Fig.2), different jet sizes (Fig.4) and
different concentration dilutions (Fig.5). Laboratory
observations [21] also confirm that different
realizations of a thermal jet have different and
unique shapes. Both numerical simulations as well
as the laboratory experiments do not produce two
identical realizations. In the following three
sections, a typical realization of the thermal jet from
the present numerical model and from [21]
laboratory observations is described to get a general
idea of the characteristics of a thermal.
ii) Protuberances
Fig.6 displays the side-view of the growth of the
thermal jet in cross flow from the present
simulation. It consists of ten consecutive pictures of
the density deficit contours at different times,
showing the development of the protuberances and
the gross movement of the thermal jet in a cross
flow. The snap shot photo of the thermal jet from
laboratory experiment [21] is shown in Fig.7. It is
striking to find that the numerical result is in close
resemblance with that from the experiment. Further
more, the simulated top-view of the thermal jet is
plotted in Fig.8, with its corresponding laboratory
shadowgraph [21] being plotted in Fig.9. Both the
results from the present simulation and [21]
experiment again display a great similarity in gross
irregular shapes.
This phenomenon of irregular shape of the thermal
jet can be explained by the fact that there exist
turbulence differences and disturbances between the
thermal jet and the ambient fluid. These differences
and disturbances will lead to protuberances as the
potential energy will do work on the disturbances.

Fig. 6. Computed side-view of the thermal jet

Z/mm

3
X/mm

Fig. 7. Side-view shadowgraph of the thermal jet


(Wong, 1991)

30

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

Z/mm

Y/mm

thermal jet illustrated in Fig.6, it may be seen that


the former is thinner than the latter. It is obviously
because the sectional contour at the center-plane
shows the bottom thickness of the horseshoe shape
concentration contour (Fig.10) rather than a
projected side view of the whole jet as represented
in Fig.6, making Fig.6 look more like the laboratory
photograph (Fig.7) than Fig.12 does.

X/mm
C
Z/mm

X/mm

Fig.8. Computed top-view of the thermal jet


Y/mm

Fig.10. Computed cross sectional concentration contours


of the thermal jet

Fig.9. Top-view shadowgraph of the thermal jet (Wong, 1991)

(a) X=129mm
Between AA and BB,

(b) X=337mm
BB and CC

Fig.11. Cross sectional concentration contours of the


thermal jet measured from the laboratory experiment
(Wong, 1991)

iii) Horseshoe shape cross section


The concentration contour distribution across five
cross sections of the thermal jet is shown in Fig.10.
Laboratory measurements [21] for two sections are
also shown in Fig.11 for comparison. Fig.10 and
Fig.11 show that the cross section of the thermal jet
has a horseshoe shape structure; the center is hollow
and the thickness along the vertical centerline is
thinner, resulting in a hollow trough along its bentover concave side of the thermal jet.

From this view it can be easily conceived that there


exists a hollow trough along the bent-over concave
side of the thermal jet. This coincides with what the
horseshoe cross sectional shape reveals. And this
phenomenon is again in agreement with the
laboratory observation illustrated in Fig.9, in which
the middle is whiter than its outer part, showing the
existence of the hollow trough.

iv) Hollow trough along its bent-over concave side


To investigate the shape of the thermal jet in greater
detail, the concentration information in the vertical
center-plane of the thermal jet is presented in Fig.12.
Comparing this figure with the side-view of the

The mechanisms governing the horseshoe shape


structure and the hollow trough of the thermal jet
are investigated through analyzing the inner flow
structure of the fluid in the following section.

31

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

h = H0

U/U0

Z/mm

h = 2 3 H0
h = 1 2 H0
h = 1 3 H0
t/T
Fig.13. Development of U-velocities at four different
heights along the vertical line at the center of the open
channel (C-C, Fig.10)

X/mm

Fig.12. Computed concentration contours in the vertical


center-plane of the thermal jet

0.1

Z/mm

3.4 Interaction between the two fluids


To investigate the mixing mechanisms of the
thermal jet and the interacting effect of the two
flowing fluids, the flow field information is
presented in Figs.13-16.
Fig.13 is the temporal development of U-velocities
at four different heights along the vertical line at the
center of the open channel (C-C section, Fig.10). It
shows that the turbulence intensity, magnitude of
velocity fluctuation in the jet body ( h = 1 2 H 0 ,
Fig.13) is stronger than in the ambient fluid

X/mm

( h = 1 H 0 , h = 2 H 0 and h = H 0 , Fig.13). It
3
3

Fig.14. Velocity field in the vertical center-plane of the


channel

is also observed that the jet has a shielding effect on


the ambient fluid ( h = H 0 , Fig.13), which causes
the velocity to slow down in the downstream of the
jet
(during
the
time
period
between

Fig.15 shows four transverse plane velocity vector


fields with their corresponding vorticity fields
1 w

with the removal of the jet, the velocity recovers to


its normal value (after 11.5

presented in Fig.16, where x = . A


2 y z
clear twin vortex flow structure can be observed at
the early stage of the jet (sections B-B and C-C,
Fig.10). This indicates that the mixing mechanism
of the thermal jet at this stage is dominated by large
scale mixing, with ambient environmental fluid
being carried and entrained from the bent-over
concave side of the jet. This results in the horseshoe
cross sectional shape of the thermal jet and the
hollow trough along its bent-over concave side. One
can also easily see that when the vortex strength is
big enough, it will result in thermal bifurcation.

t
t
10.5 and 11.5 , at h = H 0 in Fig.13). Further,
T
T
t
, h = H 0 , Fig.13).
T

Fig.14 is a temporal velocity vector field in the


vertical center-plane of the channel. It can be
clearly seen that the stream of the jet is wavy,
which is consistent with field and experimental
observations.

32

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

From Figs.15 and 16 it can also be noticed that in


the downstream of the thermal jet (sections D-D
and E-E, Fig.10), the vortex structure becomes
smaller, weaker and more uniform. This may
account for the fact that a section in the downstream
of the thermal jet has a higher value of

Dy
Dz

BB

than
Z/mm

one does in its upstream, as stated in the jet size


section.
3.5 Comparison with the conventional k- model
In contrast to the above LES results, a velocity field
and scalar contour field obtained from the
conventional k- model are plotted in Figs.17 and
18 respectively. They demonstrate only a large
scale, symmetric twin vortices and a symmetric
contour distribution. They display no variation and
no protuberance.

Z/mm

CC

4. CONCLUSION
Numerical modeling of the diffusion of a thermal
jet in flowing water is carried out using the LES
model along with the dynamic procedure.
1)
Four relationships dealing with the jet
trajectory, jet sizes and concentration
dilution are presented.
2)
The simulation shows that the thermal jet
exhibits variability among different
realizations due to turbulence differences
and disturbances in and between the two
interacting fluids.
3)
It also shows that within a realization the
thermal jet displays characteristic
protuberances, has a horseshoe cross
sectional shape and a hollow trough
along its bent-over concave side.
4)
The numerical results are in good
agreement with experimental results from
an earlier investigation.

DD

Z/mm

DD

EE
0.01m/s

EE

Z/m
m

0.01m/s

5. ACKNOWLEDGEMENTS
National Science and Engineering Research Canada
(NSERC) and Atlantic Innovation Fund (AIF)
supported this study financially, which are
gratefully acknowledged.

Y/mm

Fig.15. (continued): Velocity fields in different cross


sections (Fig.10)

33

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

5cm/s

Z/m

Z/mm

BB

X/m
Fig.17. A typical velocity vector field from the
conventional K model

Z/(cm)

Z/mm

CC

DD

X/cm
Z/mm

Fig.18. A typical scalar contour field from the


conventional K model

REFERENCES
1. Scorer RS (1957). Experiments on convection
of isolated masses of buoyant fluid. J. Fluid
Mech. 2:583-594.
2. Turner JS (1957). Buoyant vortex rings. Proc.
R. Soc. Lond. A, 239:61-75.
3. Woodward B (1959). The motion in and around
isolated thermals. Quarterly J. Royal
Meteorological Soc., Bracknell, England
85:144-151.
4. Koh RCY and Chang YC (1973). Mathematical
model for barged ocean disposal of wastes.
Pacific Northwest Envir. Res. Lab. Nat. Envir.
Res. Ctr., Corvallis, Ore.
5. Ma FX, Li CW (2001). 3D Numerical
Simulation of Ambient Discharge of Buoyant
Water. Applied Mathematical Modeling 25:375384.

Z/mm

EE

Y/mm

Fig.16. (continued): Vortex contour field in different cross


sections (Fig.10); solid and dashed lines denote clockwise
and anticlockwise vortices respectively; unit: rad/s10-2
34

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

20. Sin YS, Zhang LW (1994). Numerical


modelling of sludge dumping in a stratified
environment. Proc., Int. Conf. On Hydr. in Civ.
Engrg. 327-332.
21. Wong CF (1991). Advected Line Thermals and
Puffs. M. Phil. Thesis, Univ. of Hong Kong.

6. Li CW, Ma FX (2003). Large Eddy Simulation


of Diffusion of a Buoyancy Source in Ambient
Water. Applied Mathematical Modeling
27(8):649-663.
7. Li CW, Wang JH (2000). Large eddy
simulation of free surface shallow-water flow.
Int. J. Numer. Meth. Fluids 34:31-46.
8. Yu TS, Li CW (1998). Instantaneous discharge
of buoyant fluid in cross-flow. J. Hydr. Engrg.,
ASCE, 124(11):1161-1176.
9. Smagorinsky J (1963). General circulation
experiments with the primitive equations. Part I:
the basic experiment. Monthly Weather Review
91:91-152.
10. Lesieur M, Metais O (1996). New trends in
large-eddy simulations of turbulence. Annu. Rev.
Fluid Mech. 28:45-82.
11. Germano M, Piomelli U, Moin P, Cabot W
(1991). A dynamic subgrid-scale eddy-viscosity
model. Phys. Fluids A, 3(3):1760-65.
12. Lilly DK (1992). A proposed modification of
the Germano subgrid-scale closure method.
Phys. Fluids A, 4(3):633-635.
13. Lund TS, Ghosal S, Moin P (1993). Numerical
experiments with highly- variable eddyviscosity models. Engineering Applications to
Large Eddy Simulation, ed. U. Piomelli, S.
Ragab, 7-11, New York: ASME.
14. Liu S, Meneveau C, Katz J (1994). On the
properties of similarity subgrid-scale models as
deduced from measurements in turbulent jet. J.
Fluid Mech. 275:83-119.
15. Piomelli U (1993). High Reynolds number
calculations using the dynamic subgrid-scale
stress model. Phys. Fluids A, 5(6):1484-1490.
16. Launder BE, Spalding DB (1974). The
numerical computation of turbulent flow. Comp.
Meth. Appl. Mech. and Engrg. 3:269-289.
17. Leonard BP (1979). A stable and accurate
convective modelling procedure based on
quadratic
upstream
interpolation.
Computational Methods in Applied Mechanical
Engrg. 19:59-98.
18. Deardroff JW (1970) A numerical study of
three-dimensional turbulent channel flow at
large
Renolds numbers. J. Fluid Mech.
41:453-480.
19. Morton BR, Nguyen KC, Cresswell RW (1994).
Similarity and self-similarity in the motion of
thermals and puffs. Recent Research Advances
in the Fluid Mechanics of Turbulent Jets and
Plumes 89-116.
35

You might also like