09575820

q Institution of Chemical Engineers

Open access under CC BY-NC-ND license.

Trans IChemE, Vol 76, Part B, February 1998

MODELLING OF DISPERSION OF
TWO-PHASE RELEASES:
Part 1 Conservation Equations and Closure Relationships
M. J. PATTISON1 , R. MARTINI 2 , S. BANERJEE3 and G. F. HEWITT1 (FELLOW)
1

Department of Chemical Engineering and Chemical Technology, Imperial College, London, UK

2
CISE spa, Milan, Italy
Department of Chemical Engineering, University of California at Santa Barbara, California, USA

his paper presents a methodology for the prediction of the dispersion of a two-phase
release from a chemical or process plant. A set of one-dimensional conservation
equations for modelling such a release is derived, by taking an approach analogous to
the boundary-layer integral method, in which the instantaneous conservation equations are
averaged over a volume slice transverse to the direction of predominant plume/cloud motion.
Averaging, which makes the problem computationally tractable, removes information
regarding local gradients that govern transport of momentum, heat and mass this information
must then be supplied in the form of `closure relationships. An appropriate set of closure
relationships is then discussed, covering models for interfacial heat and mass transfer,
entrainment of air, and interactions with the wind eld. Particular features of the approach are
that it is exible enough to deal with condensation of water vapour, thermal non-equilibrium
between the phases, sloping terrain, and can handle both elevated and ground-bounded
releases. Suitable formulations for the initial and source boundary conditions, for both
instantaneous and continuous releases, are also presented.
Keywords: dispersion; two-phase releases; equations; interfacial transfer; binary droplets;
entrainment.

a manner analogous to that in the classic dam break

problem6 . This allows solution of a system of `lumped
ordinary differential equations for species mass, momentum
and energy together with an equation of state. Another
category of models is known as `similarity or `slab
models, such as that due to Colenbrander7 , and are really
extensions to plumes of the `box approach, which applies
primarily to puff releases. For transient plumes, a set of
partial differential equations involving time and one spatial
dimension as interdependent variables is solved. The third
main type of model uses a computational uid dynamics
(CFD) approach in which the equations are discretised and
solved in all three spatial dimensions (for example, Chan
et al.8 ). Although two-phase effects can be incorporated into
such codes, the CPU requirements are prohibitive at present
and make them unrealistic as a suitable tool for routine
hazard assessment.
Presented here is a new approach, which is intermediate
between the box and slab methods and the threedimensional CFD approach. The method involves averaging
the conservation equations over a volume slice transverse to
the direction of motion of the released material. This is
analogous to the boundary-layer integral method9 except
that the averaging is performed over the plume or jet rather
than the boundary layer. This type of approach has
previously been developed for individual multiphase
ows see Banerjee and Chan10 and Ishii and Mishima11 .

INTRODUCTION
In many postulated releases from chemical and process
plants, the material contained within the plant is ejected
through a break as a two-phase jet or cloud which then
disperses into the surrounding atmosphere. Estimation of
the behaviour of released material is necessary for hazard
assessment and, for two-phase releases, is obviously more
complicated than in the case of single-phase (gas) clouds.
Speci cally, the liquid phase may be dispersed as small
droplets in the ow eld, and these droplets evaporate
causing changes in the temperature and composition of the
surrounding gas. If the surrounding atmosphere is humid,
then water vapour may condense on the droplets (which
have dropped in temperature due to evaporation) and this
gives a further complicating factor in the cloud behaviour.
Different approaches, of varying complexity, have
previously been taken to predict the behaviour following
such releases, and many discussions can be found in the
literature1- 5 . The approaches can be divided into three main
categories. The rst, and simplest, is the `box model, in
which a buoyancy-driven ow in a wind eld is regarded as
being transported in a cylindrical shape whilst retaining a
self-similar internal concentration eld, i.e. the concentration distribution always has the same shape when scaled
with the cloud radius and height. The cloud can increase in
volume by entrainment of air through its boundaries, and
can slump under the action of gravity, spreading outwards in
31

32

PATTISON et al.

Figure 1. Diagram showing the path of a two-phase release into the

atmosphere.

The resulting equations can be applied to either instantaneous or continuous releases, and constitutive equations for
effects such as entrainment and interphase heat and mass
transfer can be incorporated at an appropriate level of detail.
Key features of our approach are that separate equations are
solved for the gas and liquid energy, thus dispensing with
the assumption of thermal equilibrium between the phases,
and that condensation of water vapour onto the droplets is
modelled. A prime objective of the present study was to use
this approach to develop a code (named CLOUD), which
could model the behaviour of two-phase releases and would
run at speeds of the order of real time or faster on a modern
PC or workstation.
In this paper, we present a set of averaged conservation
equations, the closure relationships required to replace
information lost in averaging, and initial and boundary
conditions for instantaneous and continuous releases. In a
subsequent paper12 , we discuss the solution procedure and
compare the results obtained from this mathematical model
with experiments.

centre-line and it is not xed in space and time; z is the

vertical co-ordinate. A diagram showing the axes used is
shown in Figure 2. The main assumption concerning s is that
the radius of curvature of the plume centre-line is large in
comparison with the plume width, i.e. ddsnWs << B1 where nWs is
the unit vector in the s-direction. It is also assumed that the
velocity of the liquid droplets and vapour/gas mixture is
equal at every point. It is straightforward to modify this to
allow droplet settling. However, in most cases considered
here the droplets are very small and hence settle very
slowly; for example, the settling velocity of 50 l m radius
droplets is about 0.2 ms- 1 (Smith et al.13). This effect is
therefore small in most practical situations, except very
close to the point of release. A discussion of the
appropriateness of this assumption is given by Kukkonen
et al.14 ).
It is considered that the volume over which the averaging
is performed is bounded by two planes (a1 and a2 ) and an
air-plume interface denoted by ai . The planes and the
interface are not, in general, xed, and can move with time.
We de ne:
uW 1 as the velocity of displacement of plane a1 ;
uW 2 as the velocity of displacement of plane a2 ;
Wv as the velocity of the material in the plume;
uW i as the velocity of displacement of the bounding surface ai .
The above velocities are functions of position and time, as is
the bounding interface.
In order to formulate suitable conservation equations, it is
necessary to make use of two theorems. The application of
the Leibniz formula (Bird et al.15) for the differentiation of
integrals to the volume in Figure 2 gives, for a general scalar
f (see Hetsroni16 , for a detailed explanation of this
procedure):

f dV
= t
t

THE CONSERVATION EQUATIONS

Figure 1 shows a schematic diagram for a typical plume.
The plume consists of a two-phase mixture of air, released
material and water and is surrounded by the external air. In
order to derive a set of one-dimensional conservation
equations for it, the plume is averaged across a crosssectional slice perpendicular to the plume centre-line and
the rate of change of a conserved quantity within this
slice is considered. The co-ordinate s lies along the plume

V (s,t)

f dV
V (s,t)

f (uWi nWi )dS

ai (s,t)

f (uW 2 nW s )dS +
a2 (s,t)

f (uW 1 nWs )dS (1)

a1 (s,t)

The application of the Gauss-Ostragradskii divergence theoW yields

rem15 to the volume slice for an arbitrary vector A
the following relationship:
=
V (s,t)

W dV
A

= s

W dV +
nW s A
V(s,t)

W dS
nW i A

(2)

ai (s,t)

The starting point for the derivation of the area-averaged

equations is to write down the local instantaneous
conservation equation for a conserved quantity w (see for
example Trusdell and Toupin17):

q w
+=
t

Figure 2. Transverse volume slice perpendicular to the plume axis.

q w vW + =

W| - q S = 0

(3)

Here W| is the ux due to molecular diffusion and S is a

source term. In the system being considered, the contribution due to W| is very small (Wallace and Hobbs18 ), but it is
included in this derivation for completeness.
Trans IChemE, Vol 76, Part B, February 1998

MODELLING OF DISPERSION OF TWO-PHASE RELEASES: PART 1

The volume and ensemble averaged form of this is then:

V (s,t)

q w dV
+
t

V (s,t)

V (s,t)

=0

q S dV

(4)

The next step is to apply equation (1) with f = q w to the

rst term in the above equation and equation (2) with
W = q w Wv + W| to the second term. This leads to:
A

V(s,t)

q w dV +
s

q w vW + W| dV

nW s
V(s,t)

V(s,t)

ai (s,t)

q SdV

nW i W| dS

q w nW i (uW i - Wv)dS ai (s,t)

q w (nW s uW 2 )dS -

q w (Wns uW 1 )dS

a2 (s,t)

(5)

a1 (s,t)

A volume average and an area average will now be de ned.

Since the curvature of the plume is small, these can be
assumed to be the same, i.e:
1
1
f dV =
f dS
(6)
f = V
A
V(s,t)

q w (nW s uW 2 )dS a2 (s,t)

q w (nW s uW1 )dS

a1 (s,t)

= d s s q
A(s,t)

w us dS

(7)

-d s

mw
dS =
ai (s,t)

(V q w ) V Wn (q w Wv W| )
+ s s
+
t

- V q S =
ai (s,t)

1
d s

q aw

dS
au

(10)

ai (s,t)

where q a is the density of the external eld, in this case the

air, and uis an entrainment velocity, equal to the velocity
across the air/plume interface.
The terms involving molecular diffusion ux will now be
dropped in order to simplify the equations. A criterion for
determining whether this assumption is reasonable is given
by Meroney19 as
u3, >

0.2g(q
q

- q a)D
a

where u, is the friction velocity and D the molecular

diffusion coef cient. Equation (9) can then be reduced to:

(A q w ) A q w (v u )
+ s
s- s
t

- Aq S =

q aw

dz
au

(11)

where the integration in the term on the right hand side is

taken over the perimeter of the volume slice.
It is convenient to de ne length scales B and H to
represent the extent of the plume in the horizontal direction
and the direction perpendicular to s and the horizontal. In
general, the edges of the plume are diffuse and there is no
uniquely de ned boundary and it is thus necessary to clarify
what is meant by these two lengths.
For the case of a free plume, the cross-stream distribution
of the dependent variables are normally taken to be
Gaussian20 . To ensure that the mass predicted to be
contained within the plume is correct, one can write:

where us is the velocity of the bounding plane in the

direction of the plume axis, i.e. us = nW s uW . Equation (5) can
now be written in the form:

q w nW i (uW i - Wv)dS
ai (s,t)

nWi W| dS + d s
s

q w us dS

(8)

A(s,t)

Dividing by d s gives:

(A q w ) A q w (v u )
Aj
+ s
s - Aq S
s- s +
t

s
1
s

The rst term on the right hand side of equation (9) can be
written as:

A(s,t)

where A is the cross-sectional area of the plume. For an

axially symmetric distribution, this assumption is correct to
rst order for non-parallel volume faces.
If the separation of the planes remains constant at a value
d s, then the last two terms in equation (5) can be expressed
by:

=d

where:
= q nWi (uW i - Wv) = rate of entrainment;
m
vs = nW s Wv = velocity of uid in s-direction;
us = nW s uW = velocity of plane in direction of plume axis.

(q w Wv + W| )dV
=

33

mw

- W| nWi dS

ai (s,t)

Trans IChemE, Vol 76, Part B, February 1998

2p rc dr
cav

(12)
2BH
where c is the pollutant concentration and cav the
concentration averaged over the cross-section; r is the
distance from the plume axis. The distribution can be
expressed as:
c = c0 exp

p r2
- 4B2

(13)

For a free plume, the assumption may be made that the

scales in both perpendicular directions are the same
(2B = H), but this does not hold for a ground-bounded
plume. In this case, if the distribution is taken to follow that
of Colenbrander7 , then the scales are de ned by:

(9)

1
= c0 c dz
0

(14)

34

PATTISON et al.

and

Table 1. Conserved quantities and source terms for governing equations.

2BHc0

Equation

c dy dz

(15)

Note that the above equations could also be applied to

other distribution functions. A discussion of concentration
pro les and effective length scales for both free and groundbounded plumes is given by Spicer and Havens21 .
Substituting A = 2BH for the area of the plume,
equation (11) can now be rewritten as:

(2BH q w ) 2BH q w (v u )
+ s

s - s
t

+ q a w a udz

= 2BH q S
(16)

This is the basic conservation equation from which the

mass, momentum and energy equations can be derived. The
form of the source terms (if any) will depend on the
particular equation being derived.
The following points should be made concerning the
application of this equation:
(1) This equation is accurate only for the case where the
curvature of the s-co-ordinate is small. Otherwise terms
arising from the rotation of the co-ordinate system would
have to be included.
(2) vs is the plume velocity in the s-direction implicitly
assumed to be in the direction of mean motion at any instant
in time and any point in space. It is a velocity in an Eulerian
frame, i.e. in a co-ordinate system stationary with respect to
the ground.
(3) Since the line of the plume axis may change with time,
the position of a point of constant s will not in general
remain xed. uW is the velocity of displacement of a point of
constant s (e.g. a bounding plane) and us is the component in
the s-direction. Naturally, for a xed coordinate system, uW
would necessarily be zero.
(4) It has been assumed that the effect of turbulence scales
are small in comparison to the scale of the plume and their
main effect is to cause entrainment of air into the plume.
This assumption is discussed in more depth by Brown22 .
The set of conservation equations used are obtained by
substituting the relevant values of w
and S into
equation (16). The appropriate expressions for these are
given in Table 1. In these equations, a is the angle between
the plume axis and the wind direction and h is the angle
between the plume axis and horizontal. v is a parameter
equal to unity for a ground-bounded plume and zero for a
free plume; values intermediate between these may be used
in the transition zone, where a plume rst touches the
ground, to avoid numerical problems. All other terms are
de ned in the nomenclature.
Physically, the source terms in the x- and y- momentum
equations arise due to the gravitational slumping of the
plume23 and the drag due to the relative motion of the plume
and ambient air. In the z-momentum equation, the gravitational slumping term is replaced by the weight of the plume
in air; it should be noted that this equation is not required for
a ground-bounded plume since the vertical position of the
plume is known. The source terms in the enthalpy equations
represent the energy transferred by solar heating, interfacial

Mass
x-momentum
y-momentum
z-momentum
Liquid energy
Gas energy
Species

1
vs cos h cos a
vs cos h sin a
vs sin h
xl h l
xg h g
mki xk

2BH q S

BH g q
- q a + s sx
s
v sin a s BH2 gq - q a + s sy
2(1 - v )gBHq - q a + s sz
9l 9 +
2BH Q
l + 2v q 9g B
9
9 Q 9l,int9 9 - mh
g9 9 +
2BH Q
g + 2v q l9 B
9
9 Q g9 ,9 int9 + mh
- 2C ki BH

v cos a

heat conduction, mass transfer and heat transfer from the

ground. Lastly, there are four species conservation equations which calculate the concentrations of liquid and
vapour phases for both water and the pollutant; a separate
conservation equation for (dry) air is not required due to the
constraint that the concentrations sum to unity.
In addition to the conservation equations, an auxillary
equation is required to determine the width and height of the
plume. Solution of the ten conservation equations yields,
amongst other parameters, the product A = 2BH, but does
not provide the individual values of B and H. For a free
plume, one can assume axial symmetry, and this leads to
2B = H, but for a ground-bounded plume, the following
relation for the slumping under gravity is used to nd B (see
Zeman24 ):
dB
gH(q - q a )
C
dt = g
q

1
2

(17)

where
1/ 2
2
(18)
1 + 4CFe B/ H
CFe is a constant, for which a value of 0.1 is recommended.
Equations of this form have been used in a number of
models for heavy gas dispersion; for example, Spicer and
Havens21 recommended the use of Cg = 1.16(q / q a )0.5 ,
based on comparisons with data on the spreading of dense
gases obtained from the Thorney Island trials.

Cg

= max 1,

THE CLOSURE RELATIONSHIPS

In order to solve the set of one-dimensional conservation
equations derived in the preceding section, relations for the
terms appearing on the right hand side of equation (16) are
required. Suitable formulations for these terms, which
correspond to entrainment, heat and mass transfer and
frictional drag, are discussed in this section.

Entrainment
Entrainment into a plume results from the effect of eddies
which arise due to the relative velocity between the plume
and its surroundings, and from ambient turbulent diffusion.
In the early stages, when the velocity of the plume is high,
the dominant mechanisms result from the relative motion of
the jet and atmosphere, but in the later stages, where the
relative velocity is small, atmospheric eddies are the
primary mechanism by which the plume disperses into the
atmosphere. The latter mechanism is often thought of as
de-entrainment of the plume into the atmosphere. In the case
Trans IChemE, Vol 76, Part B, February 1998

MODELLING OF DISPERSION OF TWO-PHASE RELEASES: PART 1

of a ground bounded plume, the boundary layer is a source
of vorticity and eddies generated there may also reach the
upper surface of the plume, causing entrainment.
For a free plume, the entrainment of air into the plume
can best be thought of as comprising three components. The
entrainment velocity, ucan thus be written in the form:
u= u1 + u2 + u3
where:

(19)

u1 is entrainment due to air/plume relative motion in axial

direction;
u2 is entrainment due to relative motion in perpendicular
direction;
u3 is entrainment due to atmospheric eddies (ambient
turbulence).
The contribution due to the rst of these terms is often
written in the form (Morton et al.25 ):
u1 = K | v - U cos h cos a |

q
q

(20)
a

where K is a constant and U is the wind velocity at the

plume centre-line. Ricou and Spalding26 conducted experiments with dense and light gas jets owing into still air and
found that the best t to their data was achieved by setting
K=0.08. Very little work has been done on two-phase jets.
That which has been done includes work by Hussain
and Segel27 on liquid jets pumped by gas bubbles and work
by Cheung and Epstein28 on entrainment into a vertical gas
bubble boundary layer. The work which is closest to the
physical situation is probably that due to MacGregor29 who
studied entrainment into spray jets. These sets of experiments suggested that the best value of K is in the range 0.1
to 0.12 and we recommend the use of K = 0.1, since this
was found by MacGregor.
Very little information appears to be available for the
entrainment arising from the cross-axis component of the
wind, so we use the expression from Ooms et al.20:
u2 = 0.5U(sin2 h + sin2 a cos2 h )
(21)
The contribution due to atmospheric turbulence can
reasonably be expected to be a function of friction velocity.
Several correlations can be found in the literature30- 32 in
which it is expressed as u3 = b u, , with the constant of
proportionality lying in the range 0.1 < b < 0.5. We used the
correlation due to Zeman24 for a neutral stability boundary
layer, which lies roughly in the middle of this range:
u3 = 0.28u,

(22)

In the case of a ground-bounded plume, entrainment may

take place through the top, the sides and the front. Most
work has dealt with entrainment through the top, since this
presents the major area available for mixing, and we shall
therefore focus on this. Data taken from entrainment
experiments are most frequently correlated with the
Richardson number, Ri, , de ned by:
Ri,

- q a )H

g(q
q

u2,

(23)

Where the mixing is generated primarily by the boundary

layer, this is the appropriate number to use. In situations
where the turbulence results mainly from velocity differences between the plume and its surroundings, a different
Trans IChemE, Vol 76, Part B, February 1998

35

Table 2. Reciprocal of Monin-Obukhov length.

Pasquil
stability class
A
B
C
D
E
F

zr = 0.01 m
0.1405
0.0848
0.0329
0.0000
0.0329
0.0848

zr

= 0.10 m
0.1109
0.0571
0.0163
0.0000
0.0163
0.0571

zr = 1.00m

zr = 3.00 m

0.0875
0.0385
0.0081
0.0000
0.0081
0.0385

0.0781
0.0319
0.0058
0.0000
0.0058
0.0319

Richardson number should be used, in which u, is replaced

by the plume/air velocity difference. A review of different
correlations is given by Fernando33 .
On the basis of comparisons of different correlations with
data on entrainment plotted by Christodoulou34 (see
Banerjee35 ), the correlations due to Colenbrander7 and
Colenbrander and Puttock36 were adopted. These correlate
the entrainment as:
u (1 + n)j
u= ,
(24)
w (Ri, )
In the above, j is the von Karman constant and n the
constant in the wind velocity pro le (U ~ zn ). w is a function
of a Richardson number and atmospheric stability:

= 0.74 + 0.25Ri0,.7 + 1.2 10- 7 Ri3,

w

= 0.74/ (1 + 0.65| Ri, | 0.6 )

Ri, > 0 or k > 0

(25a)

Ri, < 0 and k < 0

(25b)
where k is the Monin-Obukhov length. Values for this as a
function of roughness length and atmospheric stability are
given in Table 2.
However, for the case of a low Richardson number, where
ambient turbulence is the main souce of mixing, the
following expression (from Zeman24 ) should be used
instead to calculate the correct entrainment:
u= u,

0.56
0.28
w -

(26)

This is used as an upper limit on the entrainment velocity

when (25a) is used and as a lower limit where (25b) is used.
w is calculated from:
w
w

= min[1 + 5H / k , 1.5]
= min (-H/ k )- 1/ 3, 1

k >0

(27a)

k <0

(27b)

This differs from the exact form proposed by Zeman in

that we have introduced limits to avoid unphysical
predictions (for example as would happen with w > 2).
The correlations for entrainment are summarized in Table 3.
Interfacial Transfer
In the case of a two-phase release, the liquid is dispersed
within the jet or cloud in the form of small droplets, which
provides a large surface area for interfacial heat and mass
transfer. In general, the entrainment of air will cause the
droplets to evaporate, but this process is complicated by the
effect of water vapour. The droplets will act as condensation
nuclei for atmospheric water vapour and they must therefore

36

PATTISON et al.
Table 3. Correlations for entrainment.
Ground-bounded plume
Ri, > 0 or k > 0

Free plume
u= u1 + u2 + u3
u1 = K | v - U cos h cos a |

u= min(
u1 , u2 )

u2 = 0.5U(sin2 h + sin2 a cos2 h )

u2 = u,

u3 = 0.28u,

DIm
Dim

ln C

- yi,

DIm
Dim

ln C

-1

u1 =

0.56
0.28
w -

= min[1 + 5H/ k

(28)

yi is the molar fraction of species i in the vapour phase and

the subscripts d and denote conditions at the surface and
far from the droplet respectively. The subscript I denotes the
noncondensible phase (i.e., air). CT is the total molar
concentration calculated for the conditions far from the
droplet. Sh is the Sherwood number, which takes into
account the relative velocity between the droplet and air
formulations for this are given in Table 4. Dim is the
effective mixture diffusion coef cient de ned by:
yi
D
ij
1
j i
(29)
Dim = 1 - yi

u, (1 + n)j
w (Ri, )

u2 = u,

= 0.74 + 0.25Ri0,.7 + 1.2 10- 7 Ri3,

be treated as binary droplets comprising two components.

The two-component nature of the droplets has a signi cant
effect on the evaporation of the pollutant, especially in cases
where there is a strong exothermic heat of mixing with
water (for example with ammonia or hydrogen chloride)
leading to a reduction of vapour pressure, and hence,
evaporation.
Several models for the mass transfer from a binary droplet
have been derived (e.g., Newbold and Amundsen37,
Vesala and Kulmala38, Kukkonen and Vesala39 , Lage et
al.40 ). The starting point for such models has usually been the
Stefan-Maxwell transport equations (Bird et al.15) which are
then solved for a stationary, spherical droplet. The approach
taken by Newbold and Amundsen was to assume an
`effective diffusion coef cient in the solution of the
equations and this leads to the following expression for
the molar ux, Ji , of component i at the droplet surface:
yid exp
C D
= Trd Im Sh ln(C)
exp

u= max(
u1 , u2 )

u (1 n)j
u1 = , +
w (Ri, )

Ji

Ri, < 0 and k < 0

, 1.5]

w
w

0.56
0.28
w -

= 0.74/ (1 + 0.65| Ri, | 0.6 )

= min (-H / k

)- 1/ 3 , 1

where Dij is the binary diffusivity of gas i with respect to gas

j. The parameter C is de ned by:
y
C = I ,
(30)
yI ,d
More rigorous mathematical solutions of the StefanMaxwell equations exist, but comparisons by Kalkkinen41
suggest that this method should give values which are
correct to within a few percent. Comparisons with
experimental data have also shown that this approach
gives reasonable agreement42 .
The gas-side heat transfer coef cient, h, can be calculated
from the relation:
k Nu
h=
(31)
rd
where k is the thermal conductivity and Nu is the Nusselt
number, which accounts for the effect of relative motion on
the heat transfer. Table 4 gives a suitable relation for Nu.
In order to calculate the Nusselt and Sherwood numbers,
the fall velocity of droplets is needed. This requires a
knowledge of the force on a falling droplet which is given
by:
1
F = CD Ad q g uz | u|
(32)
8
where CD is the drag coef cient and u is the relative
velocity. Suitable relations for CD are provided in Table 4.
The evaporation rate is heavily dependent on the droplet
radius; for a given liquid concentration it varies approximately as the inverse square of the average radius.
Unfortunately, the droplet size is not well known (according
to a review by Herman et al.43 ) the droplets produced
from jets of superheated liquid are likely to range from
10 l m in diameter upwards, depending on the liquid used

Table 4. Correlations used for interfacial transfer.

Parameter

Source

Sherwood
number
Nusselt number

Beard and Pruppacher54

Woo and Hamielec 55
Ranz and Marshall56,57

Drag coef cient

Beard and Pruppacher58

Equation
Sh = 1 + 0.108(Re0.5 Sc0.333 )2
Sh = 0.78 + 0.308Re0.5 Sc0.333
Nu = 1 + 0.30Re0.5 Pr 0.333
24
CD =
1 + 0.102Re0.95
Re
24
CD =
1 0.115Re0.80
Re +
24
CD =
1 + 0.189Re0.632
Re

Re < 2
Re > 2
Re < 2
2 < Re < 20
20 < Re

Trans IChemE, Vol 76, Part B, February 1998

MODELLING OF DISPERSION OF TWO-PHASE RELEASES: PART 1

37

Table 5. Summary of data on droplet sizes.

Source

Droplet size
59

Brown and York

Anderson et al.44
Bettis et al.45

50 l m
100 l m
100 l m

Koestel et al.60

1676 l m

Comments
Flashing liquid jets. Freon-11 and water used
Water at superheats of 30C, 40C and 50C
Superheated Freon-11 released from ask.
Measurements made 0.5 m from source.
Theoretical work considering the shattering of water
jets at superheats of 200C

and the amount of superheat. A summary of different results

obtained for droplet sizes from ashing liquids is given in
Table 5. A value of 100 l m for the average droplet diameter
was chosen for use in the CLOUD code, based on the
ndings of Anderson et al.44 and Bettis et al.45.

The only external source of heat input into the plume

which is currently modelled is the transfer of heat between
the ground and the gas phase of the plume. The relation used
for this is taken from Colenbrander7 :
q 9g 9 = max(q91 9 , q 92 )9

(33a)

q 19 9 = 1.22 u2, q Cp (Tground - T)/ U10

(33b)

2
3

(33c)

where T is the temperature and U10 is the wind speed at an

elevation of 10 m. The term q 19 9 represents the transfer due to
thermal convection and q 19 9 that due to forced convection.
These formulations use values of u, and U10 in the ambient
air, and may not be suitable for regions very close to the
source of a jet, where the velocity may be much higher than
that of the surrounding air. In this case, the following
correlation is recommended46 :
q g9 9 = 4.184 0.0945 + 1.5577vs - 0.0899v2s + 0.003v3s

(Tground - T )

sy

= 8 CD q

sz

= 8 CD q

aU

2 3
rperp Pe

cos a cos h sin a

(36b)

aU

2 3
rperp Pe

cos a cos h sin h

(36c)

Pe is the circumference of the plume and CD is the drag

coef cient, which is taken as 0.3 (Chiang and Sill47 ).

Heat Transfer from Surroundings

2(Tground - T)2
q 92 9 = 0.089
Tground + T

(33d)

which is applicable for velocities up to about 15 ms- 1 .

The Wind Pro le

The wind pro le is usually expressed by:
U(z) =

u,

ln

z + zr
zr

-w k

(37)

where j is the von Karman constant, for which a value of

0.35 is used, zr is the roughness length, and k is the MoninObukhov length. The parameter w accounts for the effect of
stability on the pro les and is calculated as (Dyer48 ,
Paulson49 ):
5z
k >0
w =(38a)
k

= 2 ln[(1 + a)/ 2] + ln[(1 + a2)/ 2] - 2 tan- 1 a

+p /2 k <0

(38b)

where:
a

1-

16z
k

1/ 4

(38c)

The value of the Monin-Obukhov length can be obtained

from a knowledge of the roughness and the atmospheric
stability class. Values of k - 1 are given in Table 2.

Drag on the Plume

INITIAL AND BOUNDARY CONDITIONS

The wind exerts a drag on the plume through two

different mechanisms. One is by the entrainment of air into
the plume, but there is also a force due to the component of
wind in the direction perpendicular to the plume axis. If the
axes are de ned such that the x-axis is parallel to the wind
direction, then the component of wind perpendicular to the
plume axis, vperp is given by:

In situations where there is an instantaneous release, the

stored material expands rapidly, forming a cloud comprising
the vapour, liquid droplets and entrained air. Part of the
liquid may fall to the ground to form a pool. In order to set
the initial conditions, it is necessary to calculate the
composition, temperature and dimensions of the cloud
formed after this initial expansion process. Of particular
importance is a knowledge of the fraction of liquid which
immediately falls to the ground in the vicinity of the release.
On the basis of experiments involving the rupture of glass
spheres containing superheated liquid (Schmidli et al.50 ,
Schmidli51), Schmidli derived the following expression for
the fraction of liquid forming a liquid pool, fp :

vperp

= U rperp

(34)

with:
rperp = sin2 h + sin2 a cos2 h
(35)
The equations for the drag on the plume from this
mechanism are then:
1
3
Pe
s sx = CD q a U 2 rperp
(36a)
8
Trans IChemE, Vol 76, Part B, February 1998

fp

= 1 - 2.69ff 0.3 xs

(39)

where ff is the fraction of the vessel lled with liquid and xs

38

PATTISON et al.
Table 6. Formulations used for outlet ow rate.
P0

G=

Material stored as single phase gas

RT0 / Mw

G=
g
Saturated conditions in tank short pipes

Le

c +1

2c
c

2
c +1

1
2

P amb
P0

-1

= 2D

PL q

1
2

Pamb
P0

c +1
c

Equilibrium rate model

2
(hg - hl )
L
+
(Vg,0 - Vl,0 )2 T0 Cpl Le

K2

= 0.1m K = 0.61

P0
a(x0 V g + (1 - x0 )Vl )

G=g

a=

x0 (Vg,0 - Vl,0 ) +

b = a-a 1

1
2

Choked
Unchoked

Vl,0 + x0 (V g,0 - Vl,0 )

G = 2 P0 - P(T0 ) q l + G2ERM

Case where outlet is all liquid

G=

2q l (P0 - Pamb )

is the isentropic quality. A minimum value of fp = 0 is

imposed, and it is assumed that the remaining liquid is
entrained as ne droplets. The data from this set of
experiments were also used to derive an expression for the
cloud size at the end of the initial expansion process. The
velocity of expansion can be calculated from:
k2
= 2 (xs,i hg,i + (1 - xs,i )hl,i
(40)

where k is the fraction of the available energy after ashing

which is converted into kinetic energy. In the above, the
subscripts i and f signify the initial (i.e., inside the tank) and
nal states respectively. Comparisons with experimental
data suggest a value for k in the region of 5 to 9. Schmidli
recommended the use of the following equation to calculate
the duration of the expansion period, tex :
62k l rvessel
tex =
(41)
r Cp,l
The radius of the cloud after the initial expansion period can
then be calculated as:

= rvessel + uex tex

1
2

Homogeneous equilibrium model

Leung and Grolmes 64

Cp,l (Vg,0 - Vl,0 )2 T0 P0

D h0 D he

Subcooled conditions

rcloud

Unchoked

Fauske and Epstein62,63

b(1 - g a ) - log bgba -11

P0
g a (b - 1)
a(x0 Vg + (1 - x0 )Vl )
2fL
2
ab log
bg a - 1 - D
2 2
g -1
1- g b
bg - 1
2fL
log
+ bg 2 + log g = D
bg 2
b- 1

- xs,i hg,i + (1 - xs,i )hl,i

Theoretical derivation for singlephase release

See Perry et al.61

1
2

G=

u2ex

Choked
1
2

1
2

hg - hl
1
Vg - V l NT0 Cpl

G=
N

Saturated conditions in tank long pipes

P0

(RT0 / Mw )2

c +1
2(c - 1)

2
1
2

(42)

In the case of a continuous release, the main consideration

in setting the source boundary condition is the calculation of
the boundary mass ux. A number of formulae for this are

1
2

Modi ed ERM model

Fauske and Epstein62
Bernoulli type ow

available, and the choice of which is suitable is determined

by the conditions in the vessel ruptured or the breached pipe.
Suitable formulae for the calculation of the discharge rate
are presented in Table 6. One factor in the choice of
correlation is whether the pressure in the vessel is higher
than the critical pressure (normally about double the
ambient pressure), in which case the ow will be choked.
In choked ow, the pressure at the exit point is higher than
the ambient pressure, and there is then a second stage of
depressurization in which the jet undergoes an isenthalpic
decompression to atmospheric pressure (see Woodward52).
As with instantaneous releases, a proportion of the release
may fall to the ground in the vicinity of the discharge to
form a pool. Little information is available on this, though
some progress has recently been made by Woodward and
Papadourakis53 who correlated the fall-out fraction to the
size of droplets in the jet. Due to the lack of information, we
currently recommend the use of equation (39) with ff = 1.
CONCLUSIONS
A one-dimensional set of conservation equations for a
two-phase multicomponent ow has been derived by taking
area averages across planes perpendicular to the direction of
motion of the plume. A set of closure relationships
computing various appropriate models and correlations,
for application to such two-phase releases, has also been
presented and covers aspects such as heat and mass transfer,
Trans IChemE, Vol 76, Part B, February 1998

MODELLING OF DISPERSION OF TWO-PHASE RELEASES: PART 1

entrainment and frictional drag, as well as the initial/
boundary conditions. The mathematical model is now
completely formulated with the governing equations,
closure relations and boundary and initial conditions all
speci ed. The set of equations can thus be solved
numerically, and the solution methodology and comparison
with experimental data will be discussed in the next paper12 .

39

density, kg m- 3
wind drag, Nm- 1
h
angle between plume axis and horizontal
Subscripts
a
ambient air
d
droplet surface
ex
initial expansion phase
g
gas/vapour phase
I
incondensible phase (air)
0
conditions inside tank
conditions far from droplet

q
s

NOMENCLATUR E
cross-sectional area of plume, m2
droplet surface area, m2
interfacial area (Figure 2), m2
area of plume cross-section (Figure 2), m2
half width of plume, m
pollutant concentration, kg m- 3
term de ned by equation (30)
drag coef cient
total molar concentration, mol m- 3
pipe diameter, m
average of a general quantity f over plume cross-section
gravitational acceleration, ms- 2
mass ux, kg m- 2 s- 1
enthalpy, J kg- 1
height of plume, m
molecular diffusion ux
molar ux of component i at droplet surface, mol m- 2 s- 1
thermal conductivity, Jm - 1 s- 1
pipe length, m
evaporation mass transfer per unit volume, kg m- 3 s- 1
mass ux term q nWi (uWi - Wv), kg m- 2 s- 1
mass fraction of component i in phase k
molecular weight
unit vector perpendicular to interface
unit vector in s-direction
Nusselt number
pressure, Pa
perimeter of plume, m
Prandtl number
heat ux from ground, Jm- 2 s- 1
heat generation term, Jm- 3 s- 1
gas constant, J K- 1 kmol- 1
Reynolds number
Richardson number
axis along direction of plume travel, m
source terms
S
Sc
Schmidt number (kinematic viscosity/diffusion coef cient)
Sh
Sherwood number
t
time, s
uW
velocity of bounding plane, ms- 1
uW i
velocity of air/plume interface, ms- 1
U
velocity of wind eld, ms- 1
us
s-component of wind velocity, ms- 1
u1
velocity of plane a1 (Figure 2), ms- 1
u,
friction velocity, ms- 1
u
entrainment velocity, ms- 1
Wv
velocity of material in plume, ms- 1
V
speci c volume, m3 kg- 1
x
Cartesian co-ordinate, m
xk
mass fraction in phase k
xs
isentropic quality
y
Cartesian co-ordinate, m
yi
vapour mole fraction of component i
z
Cartesian co-ordinate in vertical direction, m
zr
roughness length, m
Greek letters
a
angle between plume axis and wind direction
c
ratio of speci c heats
g
critical pressure ratio
v
parameter=1 for ground-bounded plume
parameter=0 for free plume
j
von Karman constant
k
Monin-Obukhov length, m
w
conserved quantity

A
Ad
ai
a1
B
c
C
CD
CT
D
f
g
G
h
H
W|
Ji
k
L
m

m
mki
MW
nW i
nW s
Nu
P
Pe
Pr
q 9 9
9 9 9
Q
R
Re
Ri,
s

Trans IChemE, Vol 76, Part B, February 1998

REFERENCES
1. Havens, J. A., 1982, A review of mathematical models for prediction of
heavy gas atmospheric dispersion, IChemE Symposium Series No. 71,
The Assessment of Major Hazards, 124.
2. Wheatley, C. J. and Webber, D. M., 1984, Aspects of the dispersion of
denser than air vapours that are of relevance to gas cloud explosions.
Final Report EUR 9592EN for EEC Contract to UKAEA.
3. Hanna, S. R., Strimaitis, D. G. and Chang, J. C., 1991, Uncertainties in
hazardous gas model predictions, Int Conf and Workshop on Modelling
and Mitigating the Consequences of Accidental Releases of Hazardous
Materials, 2124 May, New Orleans.
4. Witlox, H. W. M. and McFarlane, K., 1994, Interfacing dispersion
models in the HGSYSTEM hazard assessment package, Atmos
Environment, 28: 29472962.
5. Hanna, S. R., Chang, J. C and Zhang, X. J., 1997, Modelling accidental
releases to the atmosphere of a dense reactive chemical (Uranium
Hexa uoride), Atmos Environment, 31: 901908.
6. Stoker, J. J., 1957, Water Waves (Pure and Applied Mathematics,
Volume 4) (Interscience Publishers, Inc., New York).
7. Colenbrander, G. W., 1980, A mathematical model for the transient
behaviour of dense vapour clouds, 3rd Int Symp Loss Prev, Basle.
8. Chan, S. T., Rodean, H. C. and Blewitt, D. N., 1987, FEM3 modelling
of ammonia and hydro uoric acid dispersion, Int Conf Vapour Cloud
Modelling (CCPS-AIChE, New York).
9. Schlichting, H., 1968, Boundary-Layer Theory (McGraw-Hill, New
York).
10. Banerjee, S. and Chan, A. M. C., 1980, Separated ow models 1.
Analysis of the averaged and local instantaneous formulations, Int J
Mult Flow, 6: 124.
11. Ishii, M. and Mishima, K., 1984, Two- uid model and hydrodynamic
constitutive relations, Nuc Eng Des, 82: 107126.
12. Pattison, M. J., Martini, R. and Banerjee, S., 1998, Modelling of the
dispersion of two-phase releases. Part 2 Numerical solution scheme
and validation, Trans IChemE, 76 (B1): 4149.
13. Smith, M. H., Park, P. M. and Consterdine, I. E., 1993, Marine aerosol
concentrations and estimated uxes over the sea, QJR Meteorol Soc,
119: 809824.
14. Kukkonen, J., Kulmala, M., Nikmo, J., Vesala, T., Webber, D. M. and
Wren, T., 1993, Aerosol cloud dispersion and the suitability of the
homogeneous equilibrium approximation, AEA/CS/HSE R1003/R
(AEA Technology, Risley, Warrington, WA3 6AT, UK).
15. Bird, R. B., Stewart, W. E. and Lightfoot, E. N., 1960, Transport
Phenomena (Wiley, New York).
16. Hetsroni, G., 1982, Handbook of Multiphase Systems (McGraw Hill
Book Company, New York).
17. Trusdell, C. and Toupin, R., 1960, The classical eld theories, In
Handbuch der Physic, 3: (Springer-Verlag, Berlin).
18. Wallace, J. M. and Hobbs, P. V., 1977, Atmospheric Science
(Academic Press Inc., Orlando).
19. Meroney, R. N., 1986, Guidelines for uid modelling of LNG vapour
dispersion. Final Report to Gas Research Institute.
20. Ooms, G., Mahieu, A. P. and Zelis, F., 1974, The plume path of vent
gases heavier than air, 1st Int Symp Loss Prev, The Hague.
21. Spicer, T. O. and Havens, J. A., 1985, Modelling the Phase I Thorney
Island experiments, J Haz Mat, 11: 237260.
22. Brown, S. T., 1987, Concentration prediction for meandering plumes
based on probability theory, Atmos Environment, 21: 13211330.
23. Gunn, D. J., 1982, Spreading and dispersion of dense vapours and
gases, IChemE Symposium Series No. 71, The Assessment of Major
Hazards, 5768.
24. Zeman, O., 1982, The dynamics and modelling of heavier-than-air,
cold gas releases, Atmos Environment, 16: 741751.
25. Morton, B. R., Taylor, G. I. and Turner, J. S., 1956, Turbulent

40

26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.

37.
38.
39.

40.
41.

42.
43.
44.
45.
46.

47.

PATTISON et al.
gravitational convection from maintained and instantaneous sources,
Proc Royal Soc, A23: 1.
Ricou, F. P. and Spalding, D. B., 1961, Measurements of entrainment
by axisymmetric turbulent jets, J Fluid Mech, 11: 2132.
Hussain, N. A., and Segel, R., 1976, Liquid jet pumped by rising gas
bubbles, J Fluids Eng, 98: 4956.
Cheung, F. B and Epstein, M., 1987, Two-phase gas bubble-liquid
boundary layer ow along vertical and inclined surfaces, Nuc Eng Des,
99: 93100.
MacGregor, S.A., 1991, Air entrainment into spray jets, Int J Heat
Fluid Flow, 12: 279283.
Tennekes, H. and Lumley, J. L., 1972, A First Course in Turbulence
(MIT Press, Boston).
Carpenter, R. J., Cleaver, R. P., Waite, P. J. and English, M. A., 1987,
The calibration of a simple model for dense gas dispersion using the
Thorney Island Phase 1 Trials data, J Haz Mat, 16: 293313.
Cleaver, R. P. and Edwards, P. D., 1990, Dispersion of a turbulent jet in
a cross ow, J Loss Prev Proc Ind, 3: 9196.
Fernando, H. J. S., 1991, Turbulent mixing in strati ed uids, Annu
Rev Fluid Mech, 23: 455493.
Christodoulou, G. C., 1986, Interfacial mixing in strati ed uids,
J Hydraul Res, 24: 77.
Banerjee, S., 1994, Entrainment into dense- uid releases. Final Report
for EC project STEP-CT91-0116.
Colenbrander, G. W. and Puttock, J. S., 1983, Dense gas dispersion
behaviour: Experimental observations and model developments,
Fourth Int Symp Loss Prev and Safety Prom Proc Ind, Harrogate,
UK, F66F75.
Newbold, F. R. and Amundsen, N. R., 1973, A model for evaporation
of a multicomponent droplet, AIChE J, 19: 2230.
Vesala, T. and Kulmala, M., 1993, Comparisons of uncoupled, lm
theoretical and exact solutions for binary droplet evaporation and
condensation, Physica A, 192: 107123.
Kukkonen, J. and Vesala, T., 1991, Aerosol dynamics in releases of
hazardous materials, Int Conf and Workshop on Modelling and
Mitigating the Consequences of Accidental Releases of Hazardous
Materials, 5385 (AIChE).
Lage, P. L. C., Hackenberg, C. M. and Rangel, R. H., 1993, Non-ideal
vaporisation of dilating binary droplets with variable properties, Int J
Heat and Mass Transfer, 36: 37313741.
Kalkkinen, J., Vesala, T. and Kulmala, M., 1991, Binary droplet
evaporation in the presence of an inert gas: An exact solution of the
Maxwell-Stefan equations, Int Comm Heat Mass Transfer, 18: 117
126.
Vesala, T., 1990, Binary droplet evaporation and condensation as
phenomenological processes, PhD Thesis (University of Helsinki).
Herman, M. W. et al., 1987, Aerosol Formation and Subsequent
Transformation and Dispersion, During Accidental Releases of
Chemicals (American Petroleum Institute, publication no. 4456).
Anderson, R. C., Erdman, C. A. and Reynolds, A. B., 1984, Droplet
size distribution from bulk ashing, Nuclear Science and Engineering,
88: 495512.
Bettis, R. J., Makhviladze, G. M. and Nolan, P. F., 1987, Expansion and
evolution of heavy gas and particulate clouds, J Haz Mat, 14: 213232.
Lin, Y. J. and Shroff, G. H., 1991, Chlorine related accidental releases
and their dispersion, Int Conf and Workshop on Modelling and
Mitigating the Consequences of Accidental Releases of Hazardous
Materials, 169182, (AIChE).
Chiang, H.-C. and Sill, B. L., 1985, Entrainment models and their
application to jets in a turbulent cross ow, Atmos Environment, 19:
pp. 1425143.

48. Dyer, A. J., 1974, A review of ux pro le relationships, BoundaryLayer Met, 7: 363372.
49. Paulson, C. A., 1970, The mathematical representation of wind speed
and temperature pro les in the unstable atmospheric boundary layer, J.
Appl Meteorol, 9: 857861.
50. Schmidli, J., Yadigaroglu, G. and Banerjee, S., 1992, Sudden release of
superheated liquids, Symposium on Basic Aspects of Two-Phase Flow
and Heat Transfer, ASME/AIChE/ANS National Heat Transfer
Conference, San Diego, California.
51. Schmidli, J., 1993, Experiments and analysis of catastrophic releases,
PhD Thesis (Swiss Federal Institute of Technology, Zurich).
52. Woodward, J. L., 1993, Expansion zone modeling of two-phase and
gas discharges, J Haz Mat, 33: 307318.
53. Woodward, J. L. and Papadourakis, A., 1995, Reassessment and reevaluation of rainout and drop size correlation for an aerosol jet, J Haz
Mat, 44: 209230.
54. Beard, K. V. and Pruppacher, H. R., 1971, A wind tunnel investigation
of the rate of evaporation of small water droplets falling at terminal
velocity in air, J Atmos Sci, 28: 14551464.
55. Woo, S. E., and Hamielec, A. E., 1971, A numerical study of
determining the rate of evaporation of small water drops falling at
terminal velocity in air, J Atmos Sci, 28: 14481454.
56. Ranz, W. E. and Marshall, W. R., 1952, Evaporation from drops, Chem
Eng Prog, 48: 141146.
57. Ranz, W. E. and Marshall, W. R., 1952, Evaporation from drops, Chem
Eng Prog, 48: 173180.
58. Beard, K. V. and Pruppacher, H. R., 1969, A determination of the
terminal velocity and drag of small water droplets by means of a wind
tunnel, J Atmos Sci, 26: 10661072.
59. Brown, R. and York, J. L., 1962, Sprays formed from ashing liquid
jets, AIChE J, 8: 149153.
60. Koestel, A. E., Gido, R. G. and Lamkin, D. E., 1980, Drop size
estimates for a loss of coolant accident, NUREG/CR-1607.
61. Perry, R. H., Green, D. W. and Maloney, J. O., 1984, Perry s Chemical
Engineers Handbook, Sixth edition, (McGraw-Hill, New York).
62. Fauske, H. K. and Epstein, M., 1987, Source term considerations in
connection with chemical accidents and vapour cloud modelling,
Int Conf Vapour Cloud Modelling, 2 4 November, 1987, Cambridge,
Mass, 251273 (AIChE).
63. Fauske, H. K. and Epstein, M., 1988, Source term considerations in
connection with vapour cloud modelling, J Loss Prev Proc Ind, 1: 75
83.
64. Leung, J. C. and Grolmes, M. A., 1987, The discharge of two-phase
ashing ow in a horizontal duct, AIChE J, 33: 524527.

ACKNOWLEDGEMENTS
This work was nanced by the European Commission under the STEP
programme under the contract STEP-CT91-0116 (DTEE); part was also
funded by CISE spa, Milan, Italy.

ADDRESS
Correspondence concerning this paper should be addressed to Dr M.J.
Pattison, Department of Chemical Engineering, University of California,
Santa Barbara, CA 93106, USA.
The manuscript was received 6 February 1997 and accepted for
publication after revision 14 August 1997.

Trans IChemE, Vol 76, Part B, February 1998