Download as pdf or txt
Download as pdf or txt
You are on page 1of 34

Pergamon Chemical Enoineering Science, Vol. 50, No. 4, pp.

695 713, 1995


Copyright 1995 Elsevier Science Lid
Printed in Great Britain. All rights reserved
0009 2509/95 $9.50 + 0.00
0009-2509(94)00246-0
MODELLING OF TWO-PHASE BLOWDOWN
FROM PIPELINES--I. A HYPERBOLIC MODEL
BASED ON VARIATIONAL PRINCIPLES
J. R. CHEN, * S. M. RI CHARDSON and G. SAVI LLE
Department of Chemical Engineering, Imperial College, London SW7 2BY, U.K.
(Received 11 October 1993; accepted in revised form 18 August 1994)
Abstract--In this paper, Geurst's variational principle for bubbly flow is extended to generalised multi-
component two-phase dispersions. The present variational principle allows both phases to be compressible
in deriving the momentum equations. A mixture energy equation is obtained using Noether's invariant
theorem and is shown to be comparable with the averaging formulation. The hyperbolicity of the equations
is achieved by forcing the flow to be marginally stable. Under the marginally stable condition, all the
information related to the structure of the flow is found to be embedded in an inertial coupling constant and
an expression for this constant is obtained based on critical flow data. The marginally stability model gives
correct sonic characteristics up to void fractions of 0.8. The clearly defined sonic characteristics make
possible the rigorous determination of the critical flow condition for rapid depressurisation of pipelines.
1. INTRODUCTION
Most acci dent s in chemical plants, nuclear power
pl ant s and offshore oil and gas pl at forms usually
result in the spillage of toxic, radi oact i ve, fl ammabl e
or explosive materials. Accurat e predi ct i on of the re-
leasing process is i mpor t ant in det ermi ni ng the conse-
quences of an accident. The predi ct ed i nformat i on
i ncl udi ng the rat e of mat eri al release, the t ot al quant -
ity released and t he physical st at e of the mat eri al is
val uabl e for eval uat i ng new process designs, process
i mprovement s and the safety of existing processes.
The bl owdown phenomenon, amongst ot her t ransi ent
release processes, is a subject of part i cul ar interest to
the chemical, oil/gas and power industries. In the
chemical industries, bl owdown of pressure vessels and
pipelines can be a hazar dous oper at i on due to the
very low t emper at ur e generat ed within t he fluid dur-
ing depressuri sat i on (Haque et al., 1990). For the
offshore oil and gas industries, cont rol l ed bl owdown
of sub-sea t r anspor t at i on lines is frequently required
in or der t o perform mai nt enance on the lines. The low
t emper at ur e generat ed duri ng the bl owdown opera-
tion may l ead to t he format i on of hydrat es and bl ock
the pipeline if free wat er is present. A detailed invest-
i gat i on i nt o the t ransi ent behavi our of pipeline blow-
down is necessary for defining oper at i onal limits
(Ellul et al., 1991). In the case of emergency bl owdown
or acci dent al r upt ur e of sub-sea pipelines, the hazard
arises not onl y because of the low t emperat ure t hat
can arise in the pi pe wall but also because of the large
t ot al efflux and high efflux rates t hat arise from the
large i nvent ory of the l ong pipelines (Ri chardson and
Saville, 1991). For example, duri ng the tragic loss of
*Author to whom correspondence should be addressed.
Present address: Institute of Applied Mechanics, National
Taiwan University, Taipei, Taiwan.
Pi per Al pha pl at form on the ni ght of 6th July 1988,
the rupt ure of three large gas lines connect ed the
Pi per Al pha platform, al t hough not the pr i mar y cause
of the accident, were act ual l y the maj or source for
escal at i on of the incident (Sylvester-Evans, 1991). The
predi ct i on of bl owdown of l ong sub-sea pipelines is
therefore of pri me i mpor t ance in the risk assessment
of offshore oil and gas installations. In the nuclear
power industries, the hypot het i cal l oss-of-cool ant ac-
cident (LOCA) is one of the most significant aspects of
the design and test of the emergency cool i ng system of
a reactor. Therefore, a fundament al st udy of the blow-
down process is crucial in the assessment of safety
practices and procedures to prevent or minimise the
consequences of cont rol l ed or uncont rol l ed releases.
Blowdown from pipes involves fast t ransi ent behav-
i our and choki ng or critical flow phenomena and is
not a trivial model l i ng task. In the simplest case of
modelling, the two phases are consi dered as a homo-
geneous (pseudo-one-phase) mixture. The result is the
homogeneous model t hat resembles the Euler equa-
tions of gas dynamics. The homogeneous model indi-
cates infinite coupl i ng between the two phases. How-
ever, the t wo-phase critical mass flow rate, one of the
most i mpor t ant factors in risk assessment, is usually
underest i mat ed by the homogeneous model (Ardron
and Furness, 1976). Ri gorous approaches based on
vari ous averagi ng met hods [e.g. Ishii (1975), Drew
(1983) and Lahey and Drew (1992)] have been appl i ed
to the local field equat i ons for each phase. This
met hod of phase averagi ng gives two sets of averaged
conservat i on equat i ons for each phases, usually called
the two-fluid model. However, the local i nformat i on
lost in the averaging process must be suppl i ed as
closure relations in or der to close the model which in
most cases is unknown. The simplest two-fluid model,
usually called the separat ed or the Wallis model
695
696 J. R. CHEN et al.
(Wallis, 1969), neglects all the i nt eract i ons between the
two phases and possesses complex characteristics, i.e.
the equat i ons are elliptic rat her t han hyperbolic, and
the model is ill-posed in the sense of Hadamar d to
initial value probl ems (Gi daspow, 1974). Apparent l y,
the model l i ng of the i nt eract i ons between the two
phases is not onl y i mpor t ant to the physical probl ems
concerned but also crucial for a pr oper mat hemat i cal
formul at i on. Despite the fact t hat an ill-posed model
of an initial value pr obl em violates the causal i t y law
(Sursock, 1982) and its sol ut i on is unstable (Drew and
Fl ahert y, 1992), stable numerical sol ut i ons of these
model s are possi bl e provi ded t hat the numeri cal diffu-
sion inherent in the numeri cal met hod, usually the
finite difference met hod, is sufficiently large to damp
the i nst abi l i t y of the model. The accuracy of the nu-
merical sol ut i on in this case is uncert ai n because mesh
refinement is not always possible. Fur t her mor e, the
complex characteristics render the ri gorous deter-
mi nat i on of boundar y condi t i ons of the flow, in par-
t i cul ar the choki ng condi t i on, difficult if not im-
possible. The existing work on t ransi ent t wo-phase
bl owdown is therefore confined to two catego-
ries: either the homogeneous model is used, e.g.
Lyczkowski e t al . (1978), Ri chardson and Saville
(1991) or an ill-posed two-fluid model is used with
unknown accuracy of finite difference appr oxi mat i on
and empi ri cal or physi cal l y unrealistic critical flow
condi t i ons, e.g. Sol bri g e t al . (1976) and Hal l e t al.
(1993), among others.
Wel l -posed model s are still possible but are usually
limited to a small range of t wo-phase flow, e.g. bubbl y
flow. For example, Lahey e t al. (1992) has devel oped
a well-posed model for bubbl y flow by t aki ng into
account all the t wo-phase i nt eract i ons t hrough a cell
model ensembl e-averagi ng met hod. However, the ex-
tension of this model to ot her flow pat t erns or high
voi d fraction regime is still far from complete. Re-
cently, an entirely different appr oach has appear ed
which uses an ext ended form of Hami l t on' s principle
to model the t wo-phase interaction. Hami l t on' s pri n-
ciple simply states t hat the mot i on of a body is a result
of mi ni mi sat i on of the time integral of the Lagrangi an
defined by the difference between kinetic energy and
pot ent i al energy. Based on the vari at i on of the
Lagrangi an, the Eul er - Lagr ange equat i ons can be
derived which can be combi ned to give the equat i ons
of mot i on of the system.
The appl i cat i on of Hami l t on' s principle in two-
phase flow probl ems was i ni t i at ed in the work of
Bedford and Drumhel l er (1978) for devel opi ng a
t heory of mixtures. The full pot ent i al of the varia-
t i onal principle in st udyi ng the i nt eract i ons or the
inertial coupl i ng in t wo-phase flow in a syst emat i c
way was first realised by Geurst (1985a, b, 1986) using
an explicit expression of averaged kinetic energy den-
sity cont ai ni ng an ext ra t erm related to the square of
relative velocity and an unknown function which he
called the virtual mass coefficient. Subsequent studies
by Wallis (I 988, 1990a) show t hat Geurst ' s vari at i onal
model is the onl y model to satisfy all the basic tests for
a pr oper macroscopi c t heory of t wo-phase flow.
Later, Pauchon and Smereka (1992) show t hat
Geurst ' s model of the equat i ons of mot i on can indeed
be recast into a form which is compat i bl e with the
convent i onal averagi ng formul at i ons except t hat all
the closure laws are det ermi ned by the unknown func-
tion. Clearly, the vari at i onal pri nci pl e can provi de an
effective way for devel opi ng and testing specific
closure laws requi red by the averaging models.
In this paper, we will concent rat e on the model l i ng
of inertial coupl i ng in t wo-phase flow using a modi -
fied vari at i onal principle based on Geur st (1985a, b,
1986). In part i cul ar, we focus on the devel opment of
the energy conservat i on equat i on and equat i ons of
mot i on for compressi bl e single- or mul t i -component
vapour - l i qui d mixtures using a t her modynami c equi-
librium assumpt i on. The charact eri st i cs of the model
are analysed. Const rai nt s of well-posedness and stable
wave pr opagat i on l ead us t o propose a class of model
which is margi nal l y stable by using an adj ust ed struc-
ture paramet er. The physical significance of this par a-
met er is discussed and an expression is also pr oposed
based on a wide range of t wo-phase critical flow and
pressure pulse pr opagat i on dat a. The result is a com-
plete set of hyperbol i c equat i ons, numerical solutions
of which for bl owdown probl ems will be present ed in
a second paper (Chen e t al . , 1993a).
2. DEVELOPMENT OF TWO- PHASE FLOW MODEL
2. 1. B a s i c d e f i n i t i o n s a n d a s s u mp t i o n s o f v a p o u r - l i q u i d
m i x t u r e s
Consi der a mul t i - component vapour - l i qui d mix-
ture in which bot h phases are allowed to be compress-
ible. The two phases can depar t from the equi l i bri um
state dependi ng on t hei r t hermal and flow states. We
shall assume here t hat the two phases are in t hermo-
dynami c equi l i bri um when the flow is homogeneous
but retain non-equi l i bri um effects from non- homo-
geneity of the flow. The t her modynami c equi l i bri um
assumpt i on will ensure the maxi mum possible mass
transfer rat e duri ng any phase change process and
significantly simplify the requi rement of model l i ng the
interracial heat / mass transfer processes into a simple
phase equi l i bri um calculation. Al t hough the neglected
t her modynami c non-equi l i bri um effects such as de-
l ayed bubbl e nucl eat i on in a superheat ed liquid can
also be consi dered under the framework considered
here, addi t i onal rel at i ons must be suppl i ed in the form
of interfacial transfer closure laws. However, there is
little i nformat i on about these closure laws except for
the simplest cases such as a st eam- wat er mixture. The
t her modynami c equi l i bri um assumpt i on is therefore
a useful and pr obabl y inevitable assumpt i on for
st udyi ng compl i cat ed t wo-phase systems such as hy-
dr ocar bon mixtures. The same assumpt i on has also
been used in ot her dynami c t wo-phase flow simula-
tion for mul t i -component hydr ocar bon mixtures, e.g.
Bendiksen e t al . (1991).
When the flow is non-homogeneous, the averaged
t her modynami c states for each phase may differ from
Modelling of two-phase blowdown from pipelines 691
each other. The non-homogeneity can be caused by
two types of dynamic effects, namely non-uniform
phase distributions and local velocity slip between the
two phases. In a one-dimensional macroscopic flow
field, the non-homogeneity is represented by the aver-
aged relative motion uo - uL, where uG and uL are the
cross-sectionally averaged velocities of the vapour
and liquid phases, respectively. By including an ex-
plicit form of flow non-homogeneity or Reynolds
stress in the kinetic energy of the two-phase system,
we shall show later that the departure from equilib-
rium can be expressed explicitly in terms of the flow
non-homogeneity or uG - uL. Regard the averaged
state of vapour as the equilibrium state. The averaged
mass densities of the vapour and liquid denoted by pG
and pL are both assumed to be functions of the aver-
aged thermodynamic pressure pc, averaged thermo-
dynamic temperature To, and individual mass frac-
tions yoi,i=r, ,N and yLi,i= ,, ,N. Any departure of
the averaged liquid state from equilibrium with the
vapour state is only caused by dynamic effects, i.e. by
the relative motion between the two phases. The
liquid is therefore attributed a dynamic temperature
TL and a dynamic pressure pL in which T, = TG and
pL = pG when uL = uo. The same equilibrium assump-
tions are also applied to other thermodynamic vari-
ables such as chemical potential pki, etc.
Let tl denote the volume fraction of the vapour
phase, or the so-called void fraction. We define the
following reduced densities of vapour and liquid
phases:
Pl = (1 - dPL> P2 = WC;. (1)
Similar expressions can be given for the entropy dens-
ities and reduced densities of component i. We have
s1 = (1 - COPLSL, s2 = WG~G, (2)
Yli = (1 - E)~LYL~, Y2i = NpGYGi,
i= l,...,N,
(3)
where sir is the averaged specific entropy of phase k.
The total internal energy density U of the mixture is
given by
u = PlhL + P2hc - EPG - (1 - ~PL
=SITL+S~TG +CPL~YI~
I
CPG~Y~~-~PG-(~ -~)PL. (4)
I
Note that the validity of the above expression requires
that the fluctuation of internal energy and other fluc-
tuations of thermodynamic properties are effectively
zero. Additional closure relations must be supplied if
these fluctuation terms are not zero and in most cases
they are not known. We therefore confine ourselves to
the above simple expression and neglect all the fluctu-
ations of thermodynamic properties. By using the
Gibbs-Duhem equation given by
1
1 yki dpki = - st dT, + m dp,, k = G or L (5)
CES 50:4-J
one gets
dU =CpLidYli + CpoidYZi + TLdS1
1 I
+ TcdS2 -(PC - PL) ~U. ( 6)
This expression is important in performing the varia-
tional analysis. Note that Geurst (1985a, b, 1986) used
a single pressure pG for both phases and the last term
in the right-hand side of eq. (6) does not appear. This
term, however, is necessary to the formulation of
a compressible liquid phase.
The interfacial mass, entropy and species transfer
rate are defined as
^
+ -(%?+F) (7)
!!!$= -(2+%) (8)
8 Y2i 8 Y2fuG
ry;dt+t= -
C?Z
i= l,...,N. (9)
Combining each equation for the two phases gives the
conservation equation of total mass, entropy and spe-
cies for non-dissipative flow:
$1 + p2) + ;(pI"L +pZUG) = o (10)
;@I + SZ) +&,uL + sZ"G) = 0 (11)
i( Yli $ Y2i) + f ( YliUL f YziUG) = 0,
i = 1, . , N. (12)
For dissipative flow, eq. (11) becomes an inequality.
Note that mass and thermal diffusions are insignifi-
cant compared to convection during the transient
process and are therefore neglected in all the dis-
cussion here. The flow will be assumed to be inviscid:
viscous or other dissipative effects will be taken into
account later by using quasi-steady-state correlations,
e.g. algebraic viscous drag laws.
When the fluid contains only one-component,
eq. (12) is no longer necessary and the Gibbs-Duhem
equation reduces to
dpk= -sxdTx+Idp,, k = G or L. (13)
pk
Equation (6) also reduces to
dU = (1 - a)pLdpL + LY~G~PG + TLdS1 + TodS2
+ (PGPG -PLPL-PC +PL)~@. (14)
2.2. Kinetic energy of two-phase flow
Following Geurst, we assume that the averaged
kinetic energy density K can be written as
K=:p,ut+:p2U~+:m(a)p1(UG-_L).
(15)
698 J. R. CHEN et al.
The first two terms represent the averaged kinetic
energy for the two phases at t hei r centres of mass in
a volume element. The last t erm t akes i nt o account
the kinetic energy associ at ed with the possi bl e velo-
city fl uct uat i on in the liquid phase due t o the pres-
ence of the vapour phase. The last t erm includes the
effects such as virtual mass accel erat i on of vapour
bubbl es having a drift velocity relative to the liquid.
For dilute non-i nt eract i ng spherical particles, it is well
known t hat m(ct)= ~t [e.g. Smereka and Mi l t on
(1991)]. Clearly, this form may not necessarily repres-
ent the correct kinetic energy due to relative mot i on in
ot her flow regimes such as annul ar or stratified flow
where the i nt eract i ons in the two phases are less
significant. However, under fast t ransi ent condi t i ons,
the two phases are st rongl y coupl ed and such flow
pat t erns usually do not occur. A similar form of kin-
etic energy is also used by Lhuillier (1985) in a differ-
ent appr oach to t wo-phase flow modelling. However,
Lhui l l i er' s appr oach ends up very si mi l ar t o t hat of
Geurst.
The coefficient (1 - ct)m(ct) is called the vi rt ual mass
coefficient by Geurst. The analysis of Smereka and
Mi l t on (1991) suggested the name Reynol ds stress
coefficient for m(~) based on the anal ogy of single-
phase t urbul ent flow. Wallis (1990a) called the func-
tion m(ct) exertia. In this work, the suggestion of
Smereka and Mi l t on (1991) will be adopt ed. One
shoul d not e t hat the above expression also neglects
the kinetic energy associ at ed with bubbl e pul sat i on
which is onl y i mpor t ant near t he bubbl e resonance
frequency. The kinetic energy associ at ed with t he ve-
locity fl uct uat i on of the vapour phase due t o the
presence of the liquid is also neglected. This is justified
because Pc is always much smaller t han PL except at
very high pressure and, in part i cul ar, in the critical
region.
2.3. Hami l t on' s pri nci pl e and equat i ons o f mot i on f o r
one- component s y s t e ms
Fi rst of all, consi der a one- component system.
Hami l t on' s pri nci pl e for a one- component system has
been st udi ed by Geurst (1985b). The present formul a-
tion is different from t hat of Geur st since the liquid is
allowed to be compressible. Nevertheless, t he equa-
tions of mot i on t urn out t o be the same. The deriva-
t i on here will be limited to one-di mensi onal flow only.
The deri vat i on of a t hree-di mensi onal form is similar
to t hat given in Geurst (1986).
The Lagrangi an densi t y of the system is written as
L = K -- U. (16)
Accordi ng to the e x t e n d e d Hami l t on' s principle
(Geurst, 1985a, b), time and space (in the present t reat -
ment the axial or z coordi nat e) i nt egrat i on of Lagran-
gian of the system is a minimum. In the vari at i ons, the
const rai nt s such as mass and ent r opy conservat i ons
must also be satisfied. The const rai nt s of mass and
ent r opy conservat i ons are i nt roduced t hr ough the use
of Lagrange multipliers. One has therefore the follow-
ing vari at i onal principle:
6 dt dz = 0 (17)
o 0
where
= L + ~o ~t( p, + p~) + ~z( p, uL + p~uc)
+ ~l - ~(S~ + S2) + (S,uL + S2uc) . ( 18)
Eul er - Lagr ange equat i ons may be obt ai ned by per-
formi ng i ndependent vari at i ons of t he dependent vari-
ables. There are two possi bl e choices of these depen-
dent variables. One can simply choose t he t hermo-
dynami c variables Pc, Tc and the flow vari abl es uc
and UL. Alternatively, one can use the combi ned vari-
ables, Pl , P2, S1, $2, UL and uc, as done by Geurst.
The former choice i nevi t abl y involves compl i cat ed
t her modynami c derivatives of densi t y and ent ropy.
Therefore, the l at t er appr oach is used. However, as
the liquid is assumed to be compressible, ~, Pc and PL,
will be used as i ndependent vari at i on variables in-
st ead of pl and P2 t o account for t he addi t i onal degree
of freedom. Also, no vari at i on of pressure, temper-
at ure and chemical pot ent i al is necessary since the
Gi bbs - Duhe m equat i on has al r eady been used. Per-
formi ng i ndependent vari at i on of ~t, PL, PC, $1, $2, UL
and uc, gives the following Eul er - Lagr ange equa-
tions:
6~: - pL u~ 1 2_
+ : pcu o [r e (a )
- (1 - ct ) m' ( g) ] pL( u~ - UL) 2 -4- ItLPL
- - Itc, Pc + Pc - - PI. + PL ( ~--t + UI. ~--Z )
- p c ( ~ + u c ~ ) = O (19a)
6p:: u~ + m( ~) ( u~ - u:) ~
- # L - - ~ - + u ~ - z = 0 (19b)
2 u G- # o - ~ - + u ~ = 0 (19c)
C
8Sl : TL + ~+UL~- ~Z = 0 (19d)
c%/ uo dt/' ] = 0 (19e)
6S2: TG + - ~ + dz /
dqJ ~tl
bUL: plUL - - pl m( ot ) ( uo - - UL) - - Pl ~Z - - $1 ~z = 0
(19f)
~uc: p2uc +p l m( ~ t ) ( u c -- uL) -- P2 - $2~ Z = 0
(19g)
Modelling of two-phase blowdown from pipelines
where the prime denotes a derivative with respect to ~c 1
+ - - ( F , - F, , sD
Substituting eqs (19b) and (19c) into (19a) gives Px
1
pL - Po = ~ pl m' ( u) ( ua - uL) 2 (20)
which relates the pressure difference between the two
phases t o the velocity difference. Also, from
eqs (19b)-(19g), it is readily shown that
T o - T L = - [ l , +m ( ~ ) p , ( ~ + l--'lIF(u-uL)21p,/lL s o - s o . _I
(21)
and
~ o - - # L = - - 2p2m( e)
+ c ' .,so _ s . . .
(22)
Equations (20)-(22) show that any departure from
t hermodynami c equilibrium is a result of dynamic
non-equilibrium. These are consistent with our pre-
vious definitions. The j umps in the averaged thermo-
dynamic states such as pressure or temperature do not
necessarily violate the local equilibrium states. For
example, the averaged temperatures for the two
phases can be different when the phase distributions
are different. This is also true when the local slip
between the two phases is zero. Even when m(~) is
zero, i.e the two phases are uncoupled, a difference in
p, T, etc. can still arise from non-uniform phase distri-
butions or slip velocity of the two phases. The only
case where the differences vanish is when the flow is
homogeneous. In this case, uo = uz and m(ct) --* oo.
Not e t hat Geurst (1985b) arrives at a slightly different
form of eq. (22) due t o a different definition of StL: the where
difference can be found t o be (Po - Pz)/P~ t hrough the
Gi bbs- Duhem equation.
The equations of mot i on can be derived from ma-
nipulation of the Eul er-Lagrange equations. Elimin-
ating the spatial derivatives in eqs (19b) and (19c)
using eqs (19d)-(19g) gives the generalised Bernoulli
theorem
( p l + p 2 ) ~ t + (S~ + S z ) ~
[ (' +l l l e o -U q =o
x l +p l m( c t ) ~ p z / d k S a - S L /
+ K+ U+ ~ p a + ( 1 - ~ ) p L = 0 . (23)
Using the pressure difference relation, eq. (20), the
above equation reduces t o the same generalised
Bernoulli theorem as obtained by Geurst (1985a, b).
Furt her manipulations yield the following equations
of mot i on in non-conservative form:
[uL -m(~)(u~ - u , . ) ] +~ ~2 u~-m(~)tuG -u~)uL
( 1 - ~ ) OpL
- m(~)(u~- u ~ ) 2 } + - -
Pl Oz
699
(24a)
~ [ u a +p,m(Ct)p= ( u o - u D]
a f l z m m ( ~ ) ( u ~ - u D u ~ }
ot Opo 1
+ (F, - F.,so)
P2 ~z P2
[ x l +p , m( ~ t ) -~ ~ # s o - s L /
(24b)
By using eq. (7), eqs (24a) and (24b) can be recast
into a form which is equivalent t o the averaged equa-
tions of Pauchon and Smereka (1992):
~
[(1 - ot)pLuL] + Oz [(1 -- t)pLu~
+ ( 1 - - OOpLm(Ot)(U o -- UL) 2]
+ (1 - ~ ) ~ z z + (1 - ~t)ptm'(~t)(u ~ - uL) 2
6(1 - ~)
x - - + FuL -- Mi a = 0 (25a)
0z
~ 2 Opo
a - t { ~ p o u d + ~ [ ~ p o u~ ] + ~ + r ~ + Mf = 0
(25b)
Mi a = ~ [(1 - . ) pLm( ~) ( uo - uL)]
+ ~z [(1 - ot)pLm(OO(U o -- UL)Uo]
+ (1 -- ~)pLm(a)(uo -- uD ~ - ; (25c)
FML = F m[ UL - - m( o t ) ( Uo - - uL)l + ( r s -- FmS L)
x [ 1 + Ol i n ( a ) ( 1 + 1 ) ] ( ~ )
--- -- FUL. (25e)
No assumption of incompressible liquid is required in
700
recasting the above equations owing t o the introduc-
tion of a separate pL. Equations (25a)-(25e) are the
generalised equations of mot i on for a one-component
two-phase flow with phase change. They are identical
to those derived by Geurst (1985b). It is therefore
concluded t hat compressibility of the liquid has no
effect on the form of the equations of motion. The
effect of interfacial mass transfer on the equations of
mot i on is uniquely determined by the above model in
eqs (25d) and (25e).
It is interesting t o compare the present model with
the conventional averaged two-fluid model. For
example, the three-dimensional ensemble-averaged
moment um equation can be written as (Lahey and
Drew, 1992):
~ p i u i
- - + V . Ot kPkUkU k -~- - - Ot k V P k + ( P k i - - p k ) V O ~ k
Ot
+ FkUki + V" ~k( Zk + ~ )
+ M~ (26)
with the moment um j ump condition given by
(puVOt k + FgUkl + M' ~) = M' [
k = C , L
J. R. CHEN e t al.
Pauchon and Smereka (1992) except t hat the term M~
is called the symmetric virtual mass acceleration.
Not e that Wallis (1990) has shown t hat M~ contains
more than just the virtual mass acceleration. Finally,
the equation of mot i on of the mixture can obtained by
adding up eqs (25a) and (25b) and replacing PL by
eq. (20):
0 2
-t[_otpauc + (1 - O O p L U L ] --}- - ~ z [ O t p G u G at- ( l - - O~)pLU 2
+ (1 - : z ) p L m( o t ) ( u a - - Uz) 2]
+ ~ + 1 ~?
~z [pL(1 -- o' ~) 2m' ( ~) ( ua - - uL) 2] = O.
(29)
2.4. E q u a t i o n s o f m o t i o n f o r m u l t i - c o m p o n e n t s y s t e m s
The extension of the above analysis t o a system
containing mul t i -component mixtures is straightfor-
ward. The only difference between a one- and
a mul t i -component system is t hat the latter has more
degrees of freedom and requires extra variations in
performing the analysis. Extra constraints of conser-
(27) vation of each species, eq. (9) must also be satisfied in
performing the variations. The Eul er-Lagrange equa-
tions are obtained by performing independent vari-
ations of ~, PL, P c , S a , $ 2 , Y l i , i = 1 . . . . . N , Y 2 i , i = 1 . . . . . N, UI..
and uc:
3ct: - - ~ u 2 1 2
2 P L L + 2 p a u c -- [m(~)
-- (1 -- c t ) m' ( ~ ) ] p L ( Uc - - UL) 2 + PC - - PL
+pL - ~+UL~z - p c ~ + u ~ =0
(30a)
b p L : i u 2 ( a ~ O ~ )
2 L W m ( o O ( t t c - - U L ) 2 - " f f ~ + U L = 0
where ~k, Tk Re, Pki , M' ~, FkUki and M 7 are the shear
stress, fluctuating moment um flux or the Reynolds
stress, interfacial pressure, interfacial force density,
interfacial moment um flux and surface tension force
density, respectively. For convenience, all the aver-
aging notations have been neglected. Compari son of
eq. (26) with (25a) and (25b) reveals t hat
Pal = PLi = P c = PL - - (1 -- ~ ) p L m' ( OO( Uc - - UL) 2
~ = - - m( o Op L( U G - - UL) 2
~ e = o 128)
"I' c = "r L = 0
M~ = - Mt = M~
F c u c i = - - F L U L i = F M C = - - F M L .
All the closure relationships are uniquely determined
and depend on known flow variables and one un-
known Reynolds stress coefficient m(~). Equat i on (28)
also shows that the interfacial pressure difference, the
velocity fluctuations and the interfacial force all have
the same origin of stress from the relative mot i on
between the two phases. The zero gas-phase Reynolds
stress is consistent with the assumption made in the
expression for the averaged kinetic energy. The van-
ishing shear stresses are also consistent with the in-
viscid assumption made earlier. The moment um j ump
condition, eq. (27), is also embedded in eq. (28) with
M~' = 0. This is expected since surface tension force is
not considered in the present framework. The above
analysis shows that the Eul er-Lagrange equations
provide not only the equations of mot i on of the sys-
tem but also the closure relations t hat completely
close the system. Similar comparisons with volume-
averaged moment um equation are also shown by
6 p 6 : j. 2 ( a t p O t p )
2 u c - - ~- +ucf f - z =0
3Sl : TL+ +u L~ z = 0
~$2: To+ - ~- +uc~ =0
6Y,,: ~L,+ \ ~ + u L ~ / = o ,
6 Y 2 i : #a~ + + u ~ c~z ] = O,
(30b)
(30c)
(30d)
(30e)
i = 1 . . . . . N
(30f)
i = 1, ... , N
(30g)
t3cp
6 U L : p l U L - - p l m ( O O ( U a - u L ) - 1 0 1
Oq O~i
- - S l - ~ , E l i = 0
7z
"7
(30h)
Modelling of two-phase blowdown from pipelines
&o
bU~: p2Ua + p l m( ~ t ) ( u ~ - - UL) - - P 2 0 ~
Orl - ~ Y2i O~` = 0. (30i)
- S2 ~ az
Substituting eqs (30b) and (30c) into (30a) gives the
same pressure difference relation, eq. (20). Also, from
eqs (30d) and (30e), it is readily shown that
T~ - - T L = ( u~ - UL)~Z. (31)
Similarly, from eqs (30f) and (30g) one gets
# O i - - # L i = ( U ~ - - U L ) ~ z , i = 1 . . . . . N. (32)
Again, eqs (31) and (32) show that any thermodyn-
amic non-equilibrium is a result of dynamic non-equi-
librium. These relations are consistent with the pre-
vious development of a one-component mixture.
However, it is no longer possible to express the tem-
perature and chemical potential differences explicitly
in terms of velocity difference. Combining eqs
(30f)-(30i) gives the following N + 1 relationships:
- - S L( T a - - T L ) , - 2 yLi (I. t , i - - # L i )
i
= [1 + m(~)] (u~ -- UL) 2
s ~ ( T ~ - T L ) + ~ y ~ ( ~ - - ~ )
i
( y ~ j - y ~ j ) [ ( u ~ - u g ) - m( ~ ) ( u ~ - u~) ~ ]
- - (yLjUo - - yaj UL)(U6 - - UL)
- - ( Y ~ j + YLJ ~ 2 ) m( a ) ( U~ - - UL)2
+ ( y Lj s a - - Y~i SL) ( Ta - - TL)
+ ~ (YLjYoi - - Yo~YLi)(IZa~ - - #U) = O,
i ~ j
j = l . . . . . N- 1 . (33c)
The generalised Bernoulli theorem for a multi-com-
ponent two-phase mixture is
(Px + P2) - ~ + (Sl +S2) + ~( Yai + Y2i)
i
+ K+ U+ ~ t p ~ + ( t - - a ) p L = 0 .
The equations of mot i on can be shown to be the same
as the one-component counterparts except that the
interfacial moment um flux terms become
r u L = F . , [ UL - - m( ~) ( u~ - - UL )] + ( r , -- r . s ~ )
x [ 1 + Pa m(a' ( 1 + -~2/dl'~](u--UL'~\ so - s---~/
701
- - ~ [ F y , - F m y L , - ( F ~ - - F m S L )
( r , - (35a)
\ s o SL / d \ UG - - UL /
F M~ = - - F , . [ u a + P l m ( a ) ( u a - u L ) - ] - ( F ~ -
(35b)
\ s~ SL / A k UG UL /
The extra effect of different volatilities of the fluids,
YG~- YL~, on the interracial moment um transfer of
a multi-component mixture appears in the last term of
FMk. Not e t hat chemical potentials cannot be ex-
pressed explicitly in terms of relative velocity but must
be solved simultaneously for all the components from
eqs (33a)-(33c).
2.5. Ene r gy cons er vat i on equat i on o f mi x t u r e s
The energy conservation equation of two-phase
mixtures can be derived from Noether' s invariant
(33a) theorem as [see Geurst (1985b) and Appendix]
~ [ p l h L + p 2 h f ; - otp(r, - (1 - a)pL + PlU 2
(33b) + z 2
2P2U6 + m(~)Px(Ua - - UL) 2]
1 2 m(~) - u L ) : ]
+ ~ z { U L [ p l h L + 2 P l U L + p t ( u a
1 2
+ u 6 [ p 2 h 6 + 2p2u6]}
+ Ozz [m(a)pl ( uo - UL) z U~] = 0. (36)
Not e that eq. (36) is valid for bot h one- and multi-
component systems. Here we are interested in the
equivalence between the present and conventional
averaging formulations. For example, the ensemble-
averaged energy conservation equation of mixtures is
(Lahey and Drew, 1992; Drew and Wallis, 1992):
1 2 eRe~
[ ap6( h~ + u 2 + e~ e) + (1 - ct)pL(hL + 2UL + L J
- ~ P G - ( 1 - ~) P L ]
(34) a
Zu2 + e~ e)
+ ~z [ a Po u G( h ~ + 2 G
& 2
+ (1 -- OOpLUL(hL + 2UL + e[e)]
- - - - [ g z ~ ' u G + (1 -- g) z~eUz - - q~
Oz
--qL g - q ~ - q [ ] = 0 (37)
where one-dimensional flow is assumed as well as that
702
there is no gravity, shear stress, internal heat source,
external heat flux and interfacial energy source from
surface tension, e~ e, qk x and q~ are the fluctuation
kinetic energy, fluctuation kinetic energy flux and
fluctuation internal energy flux, respectively. The
equivalence of these two formulations becomes more
clear if the last term on the RHS of eq. (36) is rear-
ranged as
~z [m(a) Pt (ua - uL) z ua] = ~z [m(~)pl ( u o - u t ) 2 UL]
d
+ ~Z [m(~)pl (uo -- UL)3]. (38)
One immediately infers the following closure relations
for eq. (37):
e~ e = 0
e~ e = m ( z ) p l ( u a - Uz) 2
q ~ = q~ = q ~ = 0
q~ = m( oOpl ( u ~ - - UL) 3 (39)
$Re
z = - - m( g ) p L( u c ; - - UL) 2
~g e = 0 .
These relations are consistent with the closures of the
moment um equations. These relations also reveal t hat
the last term ofeq. (36) is the summat i on of two effects:
the flux of fluctuating kinetic energy of the liquid
qL x and the product of liquid phase Reynolds stress
and mean velocity (1 - g)T~euL. Bot h terms originate
from the fluctuation of kinetic energy flux:
(u~ukuD = ( ( ( u D + u ; , p )
t t t 1 t ~ t
= (Uk)((Uk) 2 + ( ukuD) + ( u D ( u k u D + 2(UkU~Uk)
~ Tr
Re Re q ~ .
related to e k U k ' ~ k
The only surprise is t hat the factor in the definition
K
ofqr. disappears in the closure relation in eq. (39). This
is however correct and can be verified from the kinetic
energy equation.
To derive the conservation equation of kinetic en-
ergy of the mixture, we take the sum of the product of
UL with the terms in eq. (69) and the product of u~ with
the terms in eq. (70). After lengthy manipulation, one
arrives at:
i 2 1 2
-~t [ 2 p t U L + 2P2UG + m(o~)Pl (U~ - - Ul.) 2]
c3 r i . U3 _~_ p 2 U3 " ] (~
+ ~z,~,.,~ L + ~z( m( ~) m( u~ - uL)~u~
+ m( oOp~( ua - - UL)2Uo} + m' ( oOp~( ua - - UL)2~t t
dPo dPL
+ ~ u ~ - ~ z + (1 - ~)u~- f z + (u~ - UL)rU~
+ F, [u~ - u~ -- m( ~ ) ( u a - u~.)Zl = 0. (41)
J. R. CHEN et al.
One notes t hat the last term in eq. (36) also appears in
the kinetic energy equation of the mixtures. Subtract-
ing eq. (36) from eq. (41) and using the relation for
pressure difference gives
~- ~[ p2h6 ~3P6- ( 1 - ct) ~tL
+ p l h L ] - - ot - - ~-
c~ gp~
+ 7 - [ p 2 u ~ h ~ + p l U L h t ] - -
O2
OPL
- (1 - ~) Uz ~ - z - - ( U~ - - u L ) F ~
F r l u 2 I 2
- mL2 ~--2UL--m(0t)(U~--UL) 2 ] = 0 . (42)
Equat i on (42) is similar in form to the energy conser-
vation equation commonl y used in the literature, both
from averaging and phenomenological models; e.g.
Ishii (1975) and Yadigaroglu and Lahey (1976). Equa-
tion (42) is applicable t o bot h one- and multi-compon-
ent systems provided the correct relation is used for
FMa. Using the definition of ent ropy and some
lengthy algebraic manipulations, it is possible to show
that eq. (42) and the ent ropy conservation equation,
eq. (8), are equivalent. The equivalence of ent ropy and
energy equations is a result of the non-dissipative flow
assumption. However, the selection of an energy
equation rather than an ent ropy equation renders the
conventional extension of dissipative flow using
steady viscous boundar y layer approxi mat i on
straightforward.
One should note that the mixture energy conserva-
tion equation alone is enough to close the present
(40)
model where t hermodynami c equilibrium between the
two phases is assumed and the degree of non-equilib-
rium resulted from flow non-homogenei t y can be ex-
plicitly calculated from eqs (20)-(22) or eqs (31)-(33).
If, however, other t hermodynami c non-equilibrium is
present, extra information such as anot her energy
equation relating to the non-equilibrium states must
be supplied in order to close the model. In this work,
we will neglect all t hermodynami c non-equilibrium
such as delayed nucleation in superheated liquid or
delayed condensation in subcooled vapour. Duri ng
the change of state of the system, e.g. heating, the
equilibrium assumption will ensure the maximum
possible mass transfer rate while non-equilibrium
theories, provided t hat they exist, predict finite rate of
transfer. Therefore, provided that the relaxation time
of t hermodynami c equilibrium is shorter than the
relaxation time of the overall system, t hermodynami c
equilibrium assumption is justified, i.e. t h e i r r e v e r s i bl e
t r a n s f e r p r o c e s s e s c a n be we l l a p p r o x i m a t e d b y r e v e r s -
i bl e t r a n s f e r p r o c e s s e s . As will be discussed in Part II
Modelling of two-phase
of this paper (Chen et al., 1993a), these non-equi l i b-
ri um effects are onl y i mpor t ant in t he very earl y stages
of the bl owdown process and, once phase change
occurs, non-equi l i bri um states qui ckl y appr oach equi-
l i bri um ones. We also woul d like t o stress t hat , once
the t her modynami c equi l i bri um assumpt i on is made,
t here is no need t o cal cul at e F, , , Fs or Fr~ as onl y
mi xt ure species, mass and energy equat i ons are
enough t o close t he system completely. Inst ead, F' s
can be cal cul at ed from eqs (7)-(9) if necessary. Det ai l s
of t he cal cul at i on schemes are given in Chen et al.
(1993a).
In pract i cal calculations, the phase ent hal pi es and
ot her t her modynami c propert i es are det ermi ned by
equi l i bri um flash calculations. However, the liquid
ent hal py and pressure in moment um and energy
equat i ons may devi at e from equi l i bri um with the va-
pour st at e as a result of dynami c non-equi l i bri um. It is
desi rabl e t o convert these vari abl es i nt o equi l i bri um
vari abl es explicitly in the conservat i on equations. For
liquid pressure, t he pressure difference rel at i on, eq.
(20) can be readi l y used. For liquid ent hal py, some
mani pul at i ons are necessary.
Gi bbs equat i on for ent hal py densi t y can be written
a s
d( pl hL) = Tz d ( p l s z ) + u~.dPx + (1 - ~)dpL.
(43)
At equi l i bri um condi t i ons, we have
d(plh~.) = Tcd( pl s~. ) + pc dPl + (1 - ~t)dpc (44)
where the superscri pt e denot es equi l i bri um state. The
above equat i on also assumes t hat the vari at i on of
liquid densi t y is negligible. Subt ract i ng the two equa-
t i ons gives
d( pl hL) = d( pl h~) + TLd(pl SL) -- Tcd(plseL)
+ (PL -- Pc) dPl + (1 - ct)d(pL -- PC)-
(45)
Similarly, one can also derive a Gi bbs equat i on for the
ent hal py f l ux:
d(plULhL) = d(plULh[) + TL d(plULSL)
- - T c d ( p l u L S e L ) + ( P L - - I I c ) d ( p x U L )
+ (1 -- Ct)uLd(pL -- PC)" (46)
Therefore, t he mi xt ure energy equat i on, eq. (42), can
be recast as
f f _ ~ [ p E h c 63Pc . 63
+ plheL] -- ~- ~ + ~Z [-Pmuchc + paULheL]
Opt
-- [~uc + (1 -- OOUL] C3Z
= r ~ - ~( p , ( s ~ - s [ ) ) + N ( p ~u ~( s L - - ~) )
- - ( T o - - T L ) F ~ + ( ~ L - - ~ C ) r . , - - ( u c - - UL)FMC
12 12
- F,,[mUo - 2UL -- m(ot)(uc -- uz) z] (47)
blowdown from pipelines 703
where F[ is the equi l i bri um ent r opy source term de-
fined by
0
- ~ (Pl uLs~). (48) r ~ = - ~ ( p ~ s , . )
The onl y t erm unknown in eq. (47) is the non-equi l i b-
ri um liquid ent ropy. The depar t ur e from equi l i bri um
of the liquid ent r opy can be det ermi ned by the follow-
ing t her modynami c relation:
/ \
CpL
/ d ( 1 / p L) / d p L+ dTL. (49)
dSL =
The first t er m on the RHS of eq. (49) can be neglected
since the liquid is nearl y i ncompressi bl e and PL is
appr oxi mat el y constant. Therefore, the ent r opy de-
par t ur e from equi l i bri um can be appr oxi mat ed by
SL -- S~ = ASL ~ -~L ( TL -- TG). (50)
The model is therefore compl et el y closed in the frame-
work of equi l i bri um t hermodynami cs. By defining
vari ous mass and ent r opy source terms, the energy
equat i on, eq. (47), possesses a similar st ruct ure of
derivatives as the simple one-pressure Wal l i s model
(Wallis, 1969) and can be solved di rect l y using the
simplified numeri cal met hod pr oposed by Chen et al.
(1993b).
3. CHARACTERISTICS AND WAVE PROPAGATI ON
ANALYSIS
The number of charact eri st i cs associ at ed with the
model is t he same as the number of i ndependent
differential equations. Since the present model as-
sumes i sent ropi c flow, the energy equat i ons can be
decoupl ed from the mass and moment um equations.
The effect of interfacial mass transfer on the mo-
ment um equat i on for the present model is given by
eqs (25d), (25e) and eqs (35a), (35b) for single- and
mul t i -component systems, respectively. None of these
t erms cont ai ns derivatives and t hus have no effect on
the charact eri st i cs of the system. Compl et e charact er-
istics analysis of the mass and moment um equat i ons is
still very compl i cat ed because of the unknown
Reynol ds stress coefficient. The charact eri st i cs asso-
ciated with the mass conservat i on equat i ons are the
sonic charact eri st i cs which are related to the com-
pressibility of the fluid. The characteristics associ at ed
with the moment um equat i ons are the void/concentra-
tion wave or simply the voi d wave characteristics. In
the following, the sonic and voi d/ concent rat i on wave
charact eri st i cs will be anal ysed separately. Not e t hat
Geur st (1985a) has carri ed out an el abor at e character
istics analysis using linear st abi l i t y analysis.
3.1. Analysis of pressure wave propagation
For linear hyperbol i c systems, any di scont i nui t y or
shock pr opagat es onl y al ong the characteristics of the
system. For non-l i near hyperbol i c systems, the dis-
cont i nui t y does not necessarily pr opagat e al ong the
704 J. R. CHEN et al.
characteristics. Instead, any pr opagat i ng di scont i nu-
i t y is charact eri sed by the Ranki ne- Hugoni ot j ump
condi t i on where the pr opagat i on speeds usually differ
from the charact eri st i cs of the system. When the mag-
ni t ude of the di scont i nui t y becomes infinitely small,
the speeds of pr opagat i ng di scont i nui t i es appr oach
the charact eri st i c speeds of the non-l i near system.
Therefore, it is possi bl e to st udy the characteristics of
a non-l i near system by j ust l ooki ng at t he pr opaga-
tion speed of an infinitesimal per t ur bat i on wi t hout the
need of solving the compl i cat ed eigenvalue probl em.
Sergeev and Wallis (1991) were the first to utilise this
i dea to investigate small pressure pulse pr opagat i on in
a uniform t wo-phase media. Al t hough their analysis
cannot be ext ended t o the case of non-uni form flow,
the analysis provi des i mpor t ant i nformat i on regard-
ing the upper and lower bounds of inertial coupl i ng
between the two phases which is crucial in det ermi n-
ing the Reynol ds stress coefficient.
Based on the assumpt i on of i ncompressi bl e liquid,
Sergeev and Wal l i s (1991) obt ai ned the following ex-
pressi on for the pr opagat i on speed D:
(D--u____~) 2 = P 6 PL(O~2 + m ( o O ) + c t ( 1 - - 0 O P 6
\ a~ / ~PL ctp~(m(oO + 1) + (1 - ct)pLra(ct )
(5 1 )
where a6 is the speed of sound of vapour and u is the
homogeneous velocity in front of the shock. Since
PL >> P a for general cases, one gets
( D - u ~ 2 p G( c d + m(ct))
(52)
a - - T - / -
Clearly, in or der to have real and finite pr opagat i on
speed, we require t hat re(e) must be positive.
Fi gure 1 compares the experi ment al dat a on pres-
sure pul se pr opagat i on of Ruggles e t al . (1988) and
Semenov and Kost eri n (1964) with the pr opagat i on
speed predi ct ed by eq. (51) using two different func-
tion forms for m(ct): the Maxwel l i an exertia,
m(ct) = ~, which is the virtual mass coefficient of
spherical part i cl es in the dilute limit, and homogene-
ous flow, m(ct) = oo. For completeness, the speed of
sound dat a of Kar pl us (1958) are also included. The
speed of sound in t wo-phase medi a is frequency-de-
pendent and will appr oach the speed of the pressure
pulse when the sound frequency appr oaches infinity.
Fi gure 1 shows t hat the exact degree of inertial coup-
ling in di spersed t wo-phase flow lies between those
given by Maxwel l i an exertia and homogeneous flow,
i.e.
c~ < m(~) < oo (53)
with the preference given t o the Maxwel l i an exertia.
In part i cul ar, the dat a of Ruggles e t al . (1988) fit qui t e
closely to the predi ct i on using Maxwel l i an exertia.
Fi gure 1 also indicates t hat decreasing t wo-phase
coupl i ng or i nt eract i on l eads to an increase in the
wave speed. It is also interesting t o not e t hat in an-
ot her extreme case where t here is no i nt eract i on at all
between the two phases, e.g. in smoot h separat ed flow,
the pressure wave speed will bifurcate i nt o the speeds
of sound in the gas and liquid phases (Nguyen e t al.,
1981). When measuri ng the wave speed of such sys-
tems, onl y the slower sonic speed of gas is measured
0.5
0.45
0.4
K 0.35
c-
o
0.3
0
~. 0.25
..1
-u 0.2
0.15
0.1
005
Semenov, P=1.25 bar
o Karplus, P=lbar
o Ruggles, P=-lbar
- - Maxwellian Exertia
. . . . . . Homogeneous
A
I I t I
0 0.2 0.4 0,6 0.8 1
Void fraction
Fig. 1. Effect of Reynolds stress coefficient on pressure pulse propagation. The predictions are based on
a gasfliquid density ratio of 0.00139 for an air-water system.
Modelling of two-phase blowdown from pipelines
[e.g. Evans and Gous e (1970)]. I n t wo- phase bl ow-
down from pipelines, non- i nt er act i ng separ at ed flow
is rarel y encount er ed and is t herefore not consi der ed
here.
3.2. Analysis of void wave propaoation
The charact eri st i cs associ at ed wi t h t he voi d/ con-
cent r at i on wave can be st udi ed separat el y by assum-
i ng bot h phases to be i ncompr essi bl e a nd usi ng t he
concept of drift flux ( Pauchon and Smereka, 1992).
The drift flux is gi ven by
J = 0t(ua - - J0) = ~t(1 - ct)u, (54)
where j o is t he vol umet r i c flux gi ven by ctua +
( 1 - a)UL a nd u, is rel at i ve vel oci t y gi ven by ua - UL.
Us i ng t he fol l owi ng coor di nat e t r ansf or mat i ons:
z *=z - j ot , t *=t (55)
bot h mass cons er vat i on equat i ons are reduced t o
c3~ 0J
&--~ + ~z* = 0. (56)
The ki net i c energy can t herefore be expressed i n t erms
o f :t a n d J
PL p6 + pLm(Gt)
K = F( a ) J 2; r ( a ) (1 - a ) ~- a ( ~ - ~ - 2 "
(57)
The var i at i onal pri nci pl e now reads
. dt d z L = O; L=K- dp ~+~z * "
d t o d Z~
( 5 8 )
The Eul er - Lagr ange equat i ons for var i at i ons of and
J are
0
6a: F' ( a ) J 2 + ~-~ = 0 (59a)
6,I: J F + az ~ = 0. (59b)
El i mi nat i ng q~ i n eq. (59a) by usi ng eq. (59b) gives
0-~, J F - - ~ , ( r ' J 2 ) = 0. (60)
Equat i ons (56) and (60) form a reduced syst em of t he
var i at i onal model based on an i ncompr essi bl e as-
s umpt i on for bot h phases. The charact eri st i cs of t he
reduced syst em are gi ven by
J F ' _
x = ~ [ - +( r ' ~- r r ' ) ' / q. (61)
The condi t i on for t he reduced syst em to be hyper bol i c
is
2F '2 - F F " / > 0. (62)
Whe n t he equal i t y is satisfied, t he second- or der differ-
705
ent i al equat i on has t he fol l owi ng sol ut i on:
F = - e l / ( 2 + 0~) (63)
where cl and e2 are t he i nt egr at i on const ant s. By
usi ng t he def i ni t i on of F we get
c1(1 - a)ct 2 (1 - ~t)0~p~
rn(~) :t 2. (64)
(c2 + ~)PL pL
I n the di l ut e l i mi t of spherical part i cl es one knows
t hat m(a)/~ --* l / 2 as ~ ~ 0. Therefore, we choose
ci = - thpL-- Pc (65)
where th ~ 1 when ~ --* 0. rh will be called the inertial
coupling constant. The cons t ant c2 can now be deter-
mi ned by eq. (62). It t ur ns out t hat t he onl y possi bi l i t y
to satisfy t he i nequal i t y of eq. (62) is cz = 0. One can
therefore obt ai n t he condi t i on
m(~) = ~t(rh - (d~ + 2)~) (66)
i n whi ch the equal i t y in eq. (62) will be satisfied.
Equal i t y of eq. (62) also i mpl i es t hat t he t wo compl ex
conj ugat e charact eri st i cs will col l apse i nt o one real
root. The flow is therefore in a neut r al or mar gi nal l y
st abl e state. The same mar gi nal st abi l i t y condi t i on is
also obt ai ned by Geur st (1985a) usi ng l i near st abi l i t y
anal ysi s of t he compl et e var i at i onal model and by
Lhui l l i er (1985) usi ng a different t her modynami c for-
mul at i on.
When ~h = 1, eq. (66) reduces to
m(00 = ~(1 - c3o0 (67)
i n whi ch c3 >/ 3 for st abl e t wo-phase flow accor di ng to
eq. (62). Clearly, the expressi on for the Reynol ds stress
coefficient for di l ut e spherical particles or the Max-
wel l i an exertia, rn(ct)= ~, is unst abl e for any finite
voi d fraction. If t he Reynol ds stress coefficient is set to
zero, t he model reduces to t he Wal l i s model and t he
voi d wave charact eri st i cs are compl ex for any voi d
fract i on accor di ng to eq. (62).
Anot her i mpor t ant feature of t he mar gi nal st abi l i t y
condi t i on is t hat the charact eri st i cs of t he reduced
syst em i n t he Eul er i an coor di nat e become
;~ = uG (68)
whi ch i ndi cat es t hat the voi d wave pr opagat es at the
cons t ant velocity of t he gas phase and is i ndependent
of voi d fraction. It has been observed from experi-
ment s t hat voi d waves i n bubbl y flow pr opagat e at
a velocity l yi ng bet ween t he gas and l i qui d velocities
but wi t h preference for the gas velocity [e.g.
Mat uszki ewi cz et al. (1987)]. Lahey et al. (1992) also
showed t hat i nst abi l i t y of a voi d wave i n bubbl y flow
is act ual l y rel at ed to bubbl e- t o- sl ug flow regi me
t r ansi t i ons. Clearly, the exact model l i ng of a voi d
wave will be flow-regime dependent and is beyond the
cur r ent st at e-of-t he-art .
3.3. Analysis of non-linear void wave propagation
Equat i on (68) i ndi cat es the possi bl e existence of
a non- l i near voi d wave charact eri sed by t he gas velo-
706
city when the flow is margi nal l y stable. This observa-
tion led Geurst and Vreenegoor (1988) to propose the
following analysis of non-l i near void wave pr opaga-
t i on based on const ant gas velocity. This analysis
t urns out to be a conveni ent way of anal ysi ng the
margi nal st abi l i t y condi t i on of different forms of
Reynol ds stress coefficient wi t hout the need to resort
t o the characteristics analysis of the compl et e model.
We now recast their analysis in terms of the present
vari at i onal pri nci pl e in which the two phases are bot h
compressible.
The pr opagat i on equat i on of the non-l i near voi d
wave charact eri sed by const ant gas velocity and gas
density can be written as
0~ 0~
g t + u a ~zz = 0 (69)
which is reduced from the gas mass conservat i on
equat i on. The sol ut i on for a has the functional form
= f ( x - uat ) . Combi ni ng with the liquid mass con-
servat i on equat i on gives
C
uL = uG - - - (70)
1 - ~
where c is a const ant velocity.
For simplicity, it is assumed t hat the flow is isother-
mal and there is no mass transfer between the two
phases. The l at t er assumpt i on relaxes t he connect i on
between the chemical pot ent i al s of the two phases.
The addi t i onal degree of freedom is i nt roduced
t hrough the i nt roduct i on of two separat e Lagrange
mul t i pl i ers on the const rai nt s of cont i nui t y of the two
phases. The Lagrangi an becomes
FO P1 O pIU L~ FOp2 ..] OP21UG 1
=K -v + ' L Ot - 0z j "
(7 1 )
Performi ng i ndependent vari at i ons of ~, PL, Pc, u~
and u~ gives the following Eul er - Lagr ange equations:
2v t L + 2p~u~ - [rn(~)
-- (1 - ~ ) m' ( o t ) ] p ~ ( u o - - u z ) 2 + , uLpz
{ O~o2 o,~2 ~
- - p ~ , - f f ~ + u~--~-z ) = O (72a)
3PL: 2 UL2 + r n ( ~) ( u ~ - - UL) 2
--/.tz - - k, 0t + UL ~ Z ) = 0 (728)
2 u a - # a \ 0t + u a = 0 (72c)
3 u l : P l uL - p l r n ( o O( u a - UL)
Oqh
-- 01 - ~ z = 0 (72d)
J. R. CHEN et al.
Ot#2
3u~: p 2 u ~ + p t m( o ~) ( u G - uL) - p 2 ~ - z = O. (72e)
Subst i t ut i ng eqs (72b) and (72c) i nt o (72a) gives the
same pressure difference equat i on as eq. (20). The
same equat i ons of mot i on as those of Section 2.3
excluding the Fu6 and FML terms can be obt ai ned
from combi nat i ons of eqs (72b)-(72e). The cont ri bu-
tion of/~L from the liquid pressure in eq. (72b) can be
separat ed by using the Gi bbs - Duhem equation:
2 u L 2 + [-rn(00 + (1 - ~)m' (~)] (u~ - uD 2
( 0 q ,~ + U L ~ z ~ ) = 0
(73)
where the modi fi ed chemical pot ent i al kt* is a function
of gas densi t y and gas pressure. We therefore recover
the same Eul er - Lagr ange equat i on as t he formul at i on
of Geurst and Vreenegoor (1988) using an incom-
pressible liquid assumpt i on. Since Pc is const ant and
the flow is i sot hermal , Pc and therefore/~* must also
be constant. Geurst and Vreenegoor (1988) have
shown eq. (73) can onl y be satisfied provi ded t hat the
terms cont ai ni ng voi d fraction are const ant . Wi t h the
help of eq. (70), the condi t i on of existence of a non-
l i near voi d wave charact eri sed by eq. (69) can be
written as
1
[ l + m( ~ ) + ( 1 - ~ ) r n ' ( ~ ) ] ( l _ g ) 2 ( r ~+2)
(74)
where ~ is a const ant . Equat i on (74) const i t ut es a dif-
ferential equat i on for the Reynol ds stress coefficient.
Solving the differential equat i on with the condi t i on,
m(0) = 0, gives
re(g) = ~(~fi - (rh + 2)g) (75)
which is identical t o eq. (66) and implies ~ = fit. The
analysis shows that, by descri bi ng the voi d wave
pr opagat i on a pr i or i , one can also derive the corres-
pondi ng margi nal st abi l i t y condi t i on of the governi ng
equat i on wi t hout act ual l y going t hrough the det ai l ed
st abi l i t y analysis such as t hose of Section 3.2. This is
par t i cul ar l y useful in anal ysi ng the functional depend-
ence of the Reynol ds stress coefficient as shown below.
In anal ysi ng the effect of bubbl e deformat i on,
Geur st and Vreenegoor (1987) showed an interesting
result. By assumi ng t hat the Reynol ds stress coeffi-
cient is also a function of the Weber number defined
by
pL(u~ - uL) 2
We (76)
( a/ Db)
they found using linear st abi l i t y analysis t hat the
margi nal st abi l i t y condi t i on becomes
m( ~ , We ) = ~(rh - (rh + 2)cz) (77)
where rh is not a const ant but a function of
= W e ( l - ~ t ) 2 = ObpLC2/ff. Vreenegoor (1990) per-
formed a non-l i near voi d wave analysis based on eq.
Modelling of two-phase blowdown from pipelines
(69) and found t hat the Reynol ds stress coefficient also
satisfies t he same condi t i on as eq. (77).
The Weber number is of course a pl ausi bl e par a-
met er t hat will affect bubbl e deformat i on. Anot her
possi bl e effect t hat will also affect t he s t r u c t u r e of t he
flow is the viscosity and t he charact eri si ng par amet er
is the t wo-phase Reynol ds number defined by
DhpL( Uo - - uL)
R e = (78)
where Dh is t he hydraul i c di amet er of t he flow channel
and # is t he viscosity of the liquid. Therefore, the
Reynol ds stress coefficient is now defined as
m = m(ot, Re ) . Viscous di ssi pat i on at t he bubbl e sur-
face can be assumed to be negligible compar ed with
the kinetic energy of the bul k liquid at high flow
velocity. Yet the shear-i nduced bubbl e deformat i on
has a significant effect on t he fl uct uat i on kinetic en-
ergy. The inclusion of the t wo-phase Reynol ds num-
ber in the Reynol ds stress coefficient is a prel i mi nary
way of t aki ng i nt o account t he effect of viscous shear
on the bubbl e deformat i on. To show the effect of
Reynol ds number on the margi nal st abi l i t y condi t i on,
the non-l i near analysis of Geur st and Vreenegoor
(1988) is used. Performi ng i ndependent vari at i on of ct,
pL and PG, and i nt roduci ng the modi fi ed chemical
pot ent i al of liquid gives
I 2
2uL + [m + (1 -- ~ ) m~ R e mR e ] ( U6 - - uL) 2
- / ~* - ( ~ / + UL~z l ) = 0 (79)
where
0m(ct, Re ) Om( ct, Re )
( 8 0 )
m~ = Oot ' mR~ = ORe
The condi t i on for the existence of a non-l i near voi d
wave becomes
[1 + m + (1 - ct)m~ + R e mR e ] - -
( 1 - ~ ) 2
- ( ~ + 2 ) .
( 8 1 )
Under t he prescri bed voi d wave pr opagat i on, eqs (69)
and (70), the Reynol ds stress coefficient can be recast
as a function of voi d fract i on and a const ant G0:
m( ~ , R e ) = m*(ct,(o/(1 - ~t)), ~o = DhpLC/ld. (82)
Fur t her mor e,
d Go
~m* ( c t , ( o / ( l -- ct)) = m~ + ~ mR e
R e
= m~ + mRe. (83)
1 - ~ t
Therefore, eq. (81) can be rewri t t en as an or di nar y
differential equat i on:
[1 + m* + ( 1 - ~ ) d m* ] ( 1 - - 1 G t ) 2 - ~(rh + 2)
(84)
707
which is similar to eq. (74). The sol ut i on of above
equat i on satisfying m*(0) = 0 is
m* = ct(n~ - ( n~ + 2)~). (85)
Not e t hat now fit can be an ar bi t r ar y function of
Go = Re ( 1 - ct) which is a function of viscosity t hat is
not consi dered in t he inviscid flow analysis. The above
analysis suggests t hat any funct i onal dependence of
the Reynol ds stress coefficient ot her t han void frac-
t i on onl y appears in the const ant rh under the mar -
ginal st abi l i t y condi t i on eq. (66), i.e., when the flow is
margi nal l y stable; all the i nformat i on rel at i ng to the
s t r u c t u r e of the flow which is not consi dered under the
non-dissipative, inviscid flow assumpt i on is embed-
ded in the inertial coupl i ng const ant rh. This im-
por t ant feature l eads us to propose the following
margi nal st abi l i t y model.
3.4. Ma r g i n a l s t abi l i t y mo d e l
For the pur pose of model l i ng t ransi ent t wo-phase
flow in pract i cal probl ems relevant t o risk assessment,
the voi d wave characteristics are i mpor t ant but not
really essential. The most i mpor t ant ones are the sonic
charact eri st i cs which det ermi ne the critical flow rat e
and therefore must be model l ed accurately. Yet any
complex characteristics in the model will render the
model ill-posed. Therefore, it is a s s u me d t hat the voi d
wave pr opagat es at the gas velocity what ever the flow
regime. The flow regime t ransi t i ons are specified em-
pi ri cal l y by using a flow regime map. This appr oach
avoids the pr obl em of compl ex charact eri st i cs and
arrives at a flow which is margi nal l y stable for the
voi d wave characteristics. The Reynol ds stress coeffi-
cient is therefore given by the margi nal st abi l i t y con-
di t i on, eq. (66).
The margi nal st abi l i t y condi t i on for the case of
spherical particles, rh = 1, gives a negative value of
m(~t) when ct > 1/3. The negative value of Reynol ds
stress gives complex values of sonic charact eri st i cs
even t hough the voi d wave charact eri st i cs are real.
Geur st (1985a, b) suggests this br eakdown of model is
associ at ed with the br eakdown of bubbl y flow, i.e. the
flow regime is changed. This is however in cont radi c-
tion to the observat i ons of Lahey et al. (1992) t hat
flow regime t ransi t i on is rel at ed to the i nst abi l i t y of
voi d wave characteristics rat her t han the i nst abi l i t y of
sonic characteristics. Inspect i on of the margi nal st ab-
ility condi t i on, eq. (66), suggests t hat a l arger value of
rh can shift the br eakdown to a larger value of void
fraction. The analyses in previous section also show
that, when the flow is margi nal l y stable accordi ng to
the condi t i on eq. (66), all the i nformat i on relating to
the s t r u c t u r e of the flow t hat is not consi dered under
the non-dissipative, inviscid flow assumpt i on is em-
bedded in the const ant rh. It is therefore possi bl e t o
find appr opr i at e values of th for all the voi d fractions
so t hat the condi t i on of real sonic characteristics, eq.
(53), is not onl y satisfied but also gives the correct
degree of inertial coupl i ng and realistic sonic charac-
teristics. An accurat e predi ct i on of the l at t er is essen-
tial for accurat e predi ct i on of critical flow rate, one of
708 J. R. CHEN et al.
t he most i mpor t a nt factors i n t he risk assessment of
pressuri sed component s.
Unde r t he framework of i nvi sci d flow, t he onl y
var i abl e char act er i si ng t he flow st r uct ur e is t he voi d
fraction. Therefore, a dependence of th on c< is ex-
pect ed i n or der to have real soni c charact eri st i cs for
all voi d fractions. Combi ni ng t he condi t i ons, eq. (53)
wi t h eq. (66), gives
or
th - - (rh + 2)ct t> 1 (86)
1 + 2 ~
th > / - - (87)
1- - c t
A possi bl e choi ce of th is
1 + k ~ o c
rh = - - ( 8 8 )
where k~ is a cons t ant to be det er mi ned from experi-
ment s. Not e t hat eq. (88) also fulfils t he r equi r ement
t hat rh --, I when c~ ~ 0 for t he di l ut e l i mi t of spberi cal
particles.
The four charact eri st i cs of t he mar gi nal st abi l i t y
model have t he fol l owi ng forms:
2 = uG, uG, uc +_ ac. (89)
The first t wo a nd t he l ast t wo ei genval ues cor r espond
t o voi d wave charact eri st i cs and soni c characteristics,
respectively. Fo r conveni ence, denot e t he l ef t - r unni ng
soni c charact eri st i c velocity by ,IZ = uc - ac. Unde r
critical flow or choki ng condi t i ons, any di st ur bance
downs t r eam c a nnot pr opagat e upst r eam and 2Z mus t
be zero. Therefore, pr ovi ded t hat t he flow condi t i ons
at t he choki ng pl ane are known from experi ment s, t he
cons t ant kc can be det er mi ned so t hat t he cor r espond-
i ng 2Z of t he model is zero. Al t hough critical flow
dat a are a b u n d a n t i n t he l i t erat ure, critical flow dat a
suppl i ed wi t h flow condi t i ons, i n par t i cul ar the voi d
fract i on, at t he choki ng pl ane are rare. Fauske (1965)
report s t he first of these dat a for l ow-pressure
ai r - wat er systems. Mor e compr ehensi ve dat a are pr o-
vi ded by Deichsel (1988) and Deichsel and Wi nt er
(1990) for ai r - wat er syst ems at var i ous pressures be-
t ween 1.3 and 5.9 bar.
20
-20
0
-6o
?
-80
-tO0
-120
i i i
I ' I i
f i i
i i i
. . . . T l T . . . .
I i i I
~ _ x - X - X - X - X - X - X - X - X - X - X _ X x I
~ x ~ X - , I I - - X - x - x - x - x
X ~ x I I I I - X - X - x -
I I I I X - X - X - x
) I I )
i
Void fractions
0.325
0.4
0.472
0.49
- - . t 0.513
0.56
0.61
0.692
- - x - - 0.76
0. 81
[] 0.865
0.91
0.936
0.964
0.99
2 2.5 3 3.5 4 4.5 5
kc
Fig. 2. Effect of kc on the left-running sonic characteristic under critical flow conditions. Data from
Deichsel and Winter (1990) for an air-water system with P e x i t = 1.7 bara.
Modelling of two-phase
Fi gur e 2 shows t he cal cul at ed 2~- for one series of
t he dat a of Dei chsel a nd Wi nt er (1990) usi ng different
val ues of kc. It shows t hat kc -~ 3 gives t he best fit of
zero 2~- for 0 ~< ct ~< 0.8. Fi gur e 3 also shows t hat it is
i mpossi bl e to fit t he dat a to t he zero 2~- when ~t > 0.8
no mat t er what t he val ue of kc is. Fi gur e 3 shows t he
cal cul at ed 2~ for t he dat a of Faus ke (1965) a nd
Dei chsel and Wi nt er (1990). Agai n, all predi ct ed char -
acteristics are close t o zero except when ~t > 0.8. Mor e
critical flow dat a are gi ven i n Dei chsel (1988) and t he
resul t ed l ef t - r unni ng soni c charact eri st i c velocities are
pl ot t ed i n Fig. 4. Ot her t han t he dat a from t he pi pe of
0.0011 m di amet er, t he overal l agr eement wi t h zero
charact eri st i cs r emai ns t he same as Fig. 3. Dei chsel
(1988) at t r i but ed t he pecul i ar resul t s from t he
0.0011 m di amet er pi pe as a non- negl i gi bl e wall effect.
It is suspect ed t hat l i qui d is no l onger t he c ont i nuous
phase due to t he smal l pi pe di amet er. I t is therefore
suggest ed t hat t he pr esent mar gi nal st abi l i t y model
wi t h
1 +3ct
rh = (90)
1 - 0 ~
can be used for pi pes wi t h di amet er > 0.002 m up t o
= 0.8 and wi t h less accuracy for > 0.8. Geur st ' s
blowdown from pipelines 709
model of di l ut e di sper si on is therefore successfully
general i sed t o non- di l ut e t wo- phase di spersi ons.
One mus t be awar e t hat t he mar gi nal l y st abl e con-
di t i on is associ at ed wi t h t he as s umpt i on of a cons t ant
voi d wave velocity pr opagat i ng at t he drift vel oci t y of
t he vapour . Thi s is a r easonabl e appr oxi mat i on when
t he vapour phase is t he discrete phase and const i t ut e
t he char act er i si ng st r uct ur e of t he flow. Whe n t he
vapour phase is no l onger t he discrete phase and t he
l i qui d becomes t he discrete phase, e.g. l i qui d dropl et s,
t he charact eri si ng st r uct ur e of t he flow is t he discrete
l i qui d phase and t he voi d- i nduced st r uct ur e wave
shoul d pr opagat e at a velocity close to t hat of t he
l i qui d phase. Therefore, t he effect of this as s umpt i on is
most si gni fi cant when t he slip bet ween t he two phases
is very l arge a nd t he l i qui d is t he discrete phase. Thi s is
pr obabl y t he r eason why it is i mpossi bl e to fit the
critical flow dat a for ct > 0.8 no mat t er how large t he
i nert i al coupl i ng cons t ant is.
Cal cul at i ons have also been made to check t he flow
regi me of t he dat a of Dei chsel and Wi nt er (1990). It is
f ound t hat all t he dat a satisfy t he i nt er mi t t ent to
di spersed bubbl e t r ans i t i on cr i t er i on of Tai t el and
Dukl er (1976), i.e. t he flow regime is di spersed bubbl e
flow. For hi gh voi d fract i on, bubbl e flow is cert ai nl y
r.)
O
O ~
&
20
10
0
-10
-20
-30
-40
-50
-60
-70
-80
-90
-100
i I i i i i i
I I
I I I I I I I
. . . . . . . . . t - . . . . . . . ~ . . . . . . . . - t . . . . . . . . t - . . . . . . . . I - . . . . . . . ~ . . . . . . . . - t . . . . . . .
t I I I I ^ L , - ~ O I O I I
i m J e l f t f f i i
I - r 0 4 1 , I ~ l i - t I , 0
. . . . . . . . V - - - ~ . . . . - 4 . . . . . . ~ t - - . . . . . . . 4 - . . . . . . . . l - . . . . . . . . . . . . . . - ~ . . . . . . . .
t 7 I { ~ I I A t O ^ I O I
i i I I
I I I & A I [ 3
I I I OI
. . . . . . . . . ~ . . . . . . . . . . . . . . . . ~ . . . . . . . . . . . . . . . . . . . . . . . . . ~, . . . . . . . . ~ . . . . . . .
f I
. . . . . . . . . ~ . . . . . . . . . . . . . . . . ~ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 . . . . .
I I I z ~
I
It 11 . . . . . . . . I . . . . . . . . i ~ - . . . . . . .
. . . . . . . . . ,~ . . . . . . . . . . . . . . . ~, . . . . . . . . . . . . . . . . . .
I I ',
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
t I
. . . . . . . . , F . . . . . . . . . . . . . . . . . . . . . . . . . I-, . . . . . . . . . . . . . . . . . . . . . . . . . t . . . . . . . .
t I I
I
. . . . . . . . . . . . . . . . 1 I
I I
', I I
I t I t I
. . . . . . . . r . . . . . . . . . . . . . . . . - t . . . . . . . . * " . . . . . . . . r - . . . . . . . - 1 . . . . . . . . - t . . . . . . .
i
, 1 I
. . . . . . . . ~ . . . . . . . ~ . . . . . . . . ~ . . . . . . . . ~ . . . . . . . . ~ . . . . . . . ] . . . . . . . . ~ l - - r . . . .
I f I
0.2 0.3 0.4 0.5 0. 6 0.7 0.8 0. 9 1
Void fraction
D & W , * D & W , P e r D & W , P c =
Pe= 1.Tbar 1.3-2.0bar 1.8-3.5bar
a D&W, Pc= o D&W, Pe= zx Fauske,
2.3-5.2bar 2.4-5.9bar Pc= 1.156bar
Fig. 3. Left-running sonic characteristic of the marginal stability model with kc = 3 for the critical flow data
of Fauske (1965) and Deichsel and Winter (1990).
710
70
60
50
40
30
20
10
o
.,~ -10
- 2 0
O
~ -30
O
-40
~ -50
e.,
,~ - 6 0
~ -70
-80
-90
-100
-110
-120
-130
-140
J. R. CHEN e t a l .
i i i i i i d
I I r I I I
. . . . . . . . . L . . . . . . . J . . . . . . . . ~ . . . . . . . . . . . . . . . . J . . . . . . . . ~ . . . . . . . - L . . . . . . .
I I I I I I
I i I I I
I I I I I I
I I I I I I
I I I I I I
I I I I I I I
I I I I I I I
I / / / I
r . . . . . . . 7 . . . . . . . . r . . . . . . . . r . . . . . . . 1 . . . . . . . . r - - - - - - - - r . . . . . .
I I I I I I I
I I I I I I I
I I I I A I I
I I I I v I I
J . . . . . . . . . ~- . . . . . . . ~ . . . . . . . . . - - , - - - - ~ . . . . . . . , - - - , m, , , , ~. l - . . . . - ~. . . . .
I I I I I " @ I N I l 0
, , , , . ' ~ , . ~ l " l , ' . o . , . ~,
i I i I 1 7 0 U I O ~ I ~
. . . . . . . ~ . . . . . . . . ~ . . . . . . . . ~ . . . . . . . ~ . . . . . . . _ , _ - . . ~ , . _ - ~ . . . . o -
. . . . . . . . . . . . ~ I l i ~ . . . . . . . . . . ( ~ I I ~ . . . . . . . . . . . ~ I I I * - - ~ I ~ I ~ . . . . 4
, , ~ n , t 9 , , o ~ I ~ , , Q Q
r 7 ~ r ~ r o ~ f
i m e I i i 13 t ~ l n
I t I 0 I I I - I - ! I
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ~ . . . . . . . ~ -- ~ I I I ~ . . . . " I l l ~ I I I ~
I I I I L I I I / - - I I
, , a , , u , , , -, ~, ~, ~ * :
. . . . . . . . . L . . . . . . J . . . . . . . . I . . . . . . . . ~ . . . . . . . J . . . . . . . . I . . . . [
I I I I I i # I l l l l l Z~ m~ ~
I I I I I r l I - - ~ ~ - -
. . . . . . . . L . . . . . . . J . . . . . . . . l . . . . . . . . L . . . . . . . . . . . . . 1 . . . . .~ --
I l I I I I r
I i I I i i - - l u
I I I I I I I
. . . . . . . . ,~ . . . . . . . ~ . . . . . . . . t . . . . . . . . ,~ . . . . . . . fi . . . . . . . . f . . . . a - - - h T r . . . . .
I I I I I I I
I I I I I I I
. . . . . . . . r . . . . . . . 7 . . . . . . . . T . . . . . . . . f " . . . . . . . 7 . . . . . . . . T . . . . . . . . F ' I - - 1 ~ . . . .
I I I I I I
I I I I I I I
I I I I I I I
0. 2 0. 3 0. 4 0.5 0. 6 0. 7 0. 8 0. 9 1
Void fraction
a D=0. 0011 m o D=0. 0019 m * D=0. 003 m D=0. 005 m
Fig. 4. Left-running sonic characteristic of the marginal stability model with kc = 3 for the critical flow dat a
of Deichsel (1988).
n o t pos s i bl e becaus e t he l i qui d l evel is not e n o u g h t o
s u p p o r t b u b b l e di s per s i on a n d b u b b l e s s houl d co-
al esce i nt o a c o n t i n u o u s gas pha s e a n d f or m a differ-
e nt fl ow pa t t e r n, i.e. t he a nnul a r - di s pe r s e d flow. The
Tai t el a n d Du k l e r (1976) cr i t er i on f or t r a ns i t i on be-
t ween a nnul a r - di s pe r s e d a n d di s pe r s e d- bubbl e fl ow
de pe nds onl y o n t he e qui l i br i um l i qui d level. If t he
e qui l i br i um l i qui d l evel falls bel ow hal f of t he pi pe
di amet er , a nnul a r - di s pe r s e d fl ow is expect ed; i f t he
e qui l i br i um l i qui d l evel is gr e a t e r t h a n ha l f of t he pi pe
di amet er , di s pe r s e d- bubbl e fl ow is expect ed. I n t e r ms
of voi d f r act i on, t hi s cr i t er i on c o r r e s p o n d s t o t he
c ondi t i on ~ < 0.5 for di s pe r s e d- bubbl e fl ow a n d
> 0.5 for a nnul a r - di s pe r s e d flow. Ther ef or e, t he
pr es ent mode l of t wo- pha s e di s pe r s i on is st r i ct l y ap-
pl i cabl e t o ~ ~< 0.5 onl y. The fact t h a t s oni c c ha r a c t e r -
i st i cs ar e pr edi ct ed well up t o ~ = 0.8 suggest s t h a t
l ar ge di s t or t e d b u b b l e fl ow or sl ug fl ow mi g h t still
per s i s t at voi d f r act i ons gr e a t e r t h a n 0.5 u n d e r cr i t i cal
fl ow c ondi t i on. F u r t h e r s t udi es o n t he di s per s ed-
b u b b l e t o a nnul a r - di s pe r s e d t r a ns i t i on mi ght hel p t o
cl ari fy s uch a possi bi l i t y.
I t is al so a p p a r e n t t h a t t he pr es s ur e di f f er ence t e r m
gi ven by eq. (20) be c ome s negat i ve wh e n > 0.5 u n d e r
t he ma r gi na l s t abi l i t y condi t i on. Thi s unphys i c a l re-
sul t r esul t s f r om t he a s s u mp t i o n of voi d wave p r o p a -
ga t i on at a c o n s t a n t ga s - pha s e vel oci t y a l t h o u g h t he
negat i ve pr es s ur e di f f er ence does not l ead t o a c a t a -
s t r ophi c f ai l ur e of t he mode l wh e n ~ > 0.5. Thi s al s o
coi nci des wi t h t he i deal Ta i t e l - Du k l e r t r ans i -
t i on b o u n d a r y of di s pe r s e d- bubbl e / a nnul a r - di s -
per s ed flow. Cl ear l y, a p r o p e r de s c r i pt i on of voi d
wave p r o p a g a t i o n is essent i al t o f or mul a t e t he fl ow-
r egi me d e p e n d e n t t wo- f l ui d model . Howe ve r , t o
ma t c h voi d wave p r o p a g a t i o n at t he t wo ext r emes of
voi d f r act i on s i mul t a ne ous l y is not a t r i vi al t a s k a n d
does n o t necessar i l y yi el d c ons i s t e nt r es ul t s wi t h t he
s i ngl e- par t i cl e l i mi t . F o r t hes e r eas ons , t hi s wo r k will
be c o n t e n t wi t h t he pr e s e nt s i mpl e a s s u mp t i o n s of
voi d wave pr opa ga t i on.
4 . C O N C L U S I O N
The i ner t i al l y c oupl e d e q u a t i o n s of mo t i o n of
b u b b l y f l ow of Ge u r s t (1985a, b) ar e e xt e nde d t o gen-
er al i s ed t wo- pha s e di sper si ons. The e xt e ns i on c a n be
s u mma r i s e d as fol l ows.
The modi f i ed va r i a t i ona l pr i nci pl e al l ows b o t h
pha s e s t o be compr es s i bl e.
An ener gy e q u a t i o n of t he mi xt ur e is o b t a i n e d by
us i ng No e t h e r ' s i n v a r i a n t t h e o r e m a n d is s h o wn
t o be c o mp a r a b l e wi t h t he a ve r a gi ng f or mul a -
t i ons.
Modelling of two-phase blowdown from
The cl osure of i nt erfaci al mass and energy t r ans- Fm
fer is est abl i shed by usi ng a t her modynami c FMk
equi l i br i um assumpt i on. F~
Hyper bol i ci t y of t he equat i ons is achi eved by Fr i
forci ng t he flow to be mar gi nal l y stable. Un d e r 2
t he mar gi nal st abi l i t y condi t i on, all t he i nf or ma- #
t i on rel at ed t o t he st r uct ur e of the flow t hat is not #k
consi der ed under t he framework of non- di ssi - P~.j=I.2
pat i ve, i nvi sci d flow is f ound to be embedded i n
an i nert i al coupl i ng const ant . ,Ok
An expressi on for t he i nert i al coupl i ng cons t ant r
is obt ai ned based on critical flow dat a. The mar - Zk
Re
gi nal st abi l i t y model gives correct soni c charac- r~
teristics up to ct = 0.8.
The wel l -posedness a nd cl earl y defi ned char act er -
istics of t he mar gi nal st abi l i t y model make t he numer -
ical s i mul at i on of physi cal pr obl ems possi bl e and
meani ngful . I n t he second par t of this paper, t he
simplified finite difference met hod of Chen et al.
(1993b) for equi l i br i um t wo- phase flow model s will be
ext ended and appl i ed to t he mar gi nal st abi l i t y model
for t he cases of bl owdown from pipes cont ai ni ng
mul t i - component mi xt ures.
Acknowledgements--The authors would like to thank Prof.
J. A. Geurst who kindly read and commented on a prelimi-
nary version of this paper. Financial support from British
Gas plc for JRC through the award of research scholarship is
also gratefully acknowledged.
NOTATION
a~ speed of . sound of vapour
c i nt egr at i on cons t ant
D pr opagat i on speed of pressure pul se
Se f l uct uat i on ki net i c energy of phase k
e k
hk specific ent hai py of phase k
J drift flux
K ki net i c energy densi t y
L Lagr angi an densi t y
m Reynol ds stress coefficient
th i nert i al coupl i ng cons t ant
M~ interfacial force densi t y
M~' surface t ensi on force densi t y
N numbe r of c ompone nt of t he fluid
Pk pressure of phase k
qk r f l uct uat i on ki net i c energy flux of phase k
qk r f l uct uat i on i nt er nal energy flux of phase k
sk specific ent r opy of phase k
t t i me
Tk t emper at ur e of phase k
Uk velocity of phase k
U t ot al i nt er nal energy densi t y
Yk, mass fract i on of c ompone nt i i n phase k
Yo.j=I.2 reduced densities of l i qui d and vapour
phases of c ompone nt i
z axial coor di nat e
Greek l et t ers
ct void fract i on
pipelines 711
interfacial mass t ransfer rat e
interfacial mo me n t u m flux of phase k
interracial equi l i br i um ent r opy t ransfer rat e
interfacial species t ransfer rat e
charact eri st i cs or ei genval ues of t he model
viscosity of l i qui d
chemi cal pot ent i al of phase k
reduced densities of l i qui d and vapour
phases
densi t y of phase k
surface t ensi on
shear stress of phase k
f l uct uat i ng mo me n t u m flux of phase k
REFERENCES
Ardron, K. H. and Furness, R. A., 1976, A study of the
critical flow models used in reactor blowdown analysis.
Nucl. Engng Des. 39, 257-266.
Bedford, A. and Drumheller, D. S., 1978, A variational
theory of immiscible mixtures. Arch. Rat. Mech. Anal. 68,
37-51.
Bendiksen, K. O., Malnes, D., Moe, R. and Nuland, S., 1991,
The dynamic two-fluid model OLGA: theory and applica-
tion. SPE Prod. Engng 171-180.
Chen, J. R., Richardson, S. M. and Saville, G., 1993a, Model-
ling of two-phase blowdown from pipelines--II. A simpli-
fied numerical method for multi-component mixtures.
Chem. Engng Sci. (submitted).
Chen, J. R., Richardson, S. M. and Saville, G., 1993b, A sim-
plified numerical method for transient two-phase pipe
flow. Trans. Inst. chem. Engr 71A, 304-306.
Deichsel, M, 1988, Experimentelle und Analytische Unter-
suchung Adiabater Kritischer Wasser/Luft-Zweiphasen-
strrmungen in Rohren kleiner Durchmesser. Dr.-Ing. Dis-
sertation, Technische Universit/it Mfinchen.
Deichsel, M. and Winter E. R. F., 1990, Adiabatic two-phase
pipe flow of air-water mixtures under critical flow condi-
tions. Int. J. Multiphase Flow 16, 391-406.
Drew, D. A., 1983, Mathematical modeling of two-phase
flow. Ann. Rev. Fluid Mech. 15, 261-291.
Drew, D. A. and Flaherty J. E., 1992, Analysis of multiphase
flow. Proceeding of the 3rd International Workshop on
Two-Phase Flow Fund., London.
Drew, D. A. and Wallis, G. B., 1992, Fundamentals of two-
phase flow modelling. Proceeding of the 3rd International
Workshop on Two-Phase Flow Fund., London.
Ellul., I. R., King, P. E., Findly, W. A. and Delacroix, M. P.,
1991, The use of dynamic simulation in offshore multi-
phase pipeline design, in Multi-Phase Production (Edited
by A. P. Burns). Elsevier, London.
Evans, R. G. and Gouse, S. W., 1970, Pressure wave propa-
gation in adiabatic slug-annular-mist two-phase
gas-liquid flow. Chem. Engng Sci. 25, 569-582.
Fauske, H. K., 1965, Two-Phase and one-component critical
flow. Proceeding of the Symposium on Two-Phase Flow,
Exeter, England.
Geurst, J. A., 1985a, Virtual mass in two-phase bubbly flow.
Physica 129A, 233-261.
Geurst, J. A., 1985b, Two-fluid hydrodynamics of bubbly
liquid/vapour mixture including phase change. Philips d.
Res. 40, 352-374.
Geurst, J. A., 1986, Variational principles and two-fluid
hydrodynamics of bubbly liquid/gas mixtures. Physica
135A, 455-486.
Geurst, J. A. and Vreenegoor, A. J. N., 1987, Variational
approach to bubble deformation in two-phase flow. Pro-
ceedings of the 1st International Conference Ind. Appl.
Math. ( I CI AM 87), Contributions from the Netherlands,
Paris-La Villette.
Geurst, J. A. and Vreenegoor, A. J. N., 1988, Nonlinear
712 J. R. CHEN e t al.
void-fraction waves in two-phase bubbl y flow. Z A M P 39,
376-386.
Gidaspow, D., 1974, Modeling of two-phase flow, round
table discussions. Pr oceedi ng o f t he 5t h I n t e r n a t i o n a l He a t
Tr ans f er Conf erence, Tokyo.
Hall, A. R. W., Butcher, G. R. and Teh, C. E., 1993. Transient
simulation of two-phase hydrocarbon flows in pipelines.
Pr oceedi ng o f t he Eur ope an Two - Ph a s e Fl o w Group Me e t -
ing, Hannover, June 6-10.
Haque, M. A., Richardson, S. M., Saville, G. and Chamber-
lain, G., 1990, Rapid depressurisation of pressure vessels.
d. Lo s s Prey. Pr oc e s s I nd. 3, 4-7.
Ishii, M., 1975, Ther mo- f l ui d Dy nami c The or y o f Two- phas e
Fl ow. Eyrolles, Paris.
Karplus, H. B., 1958, The velocity of sound in a liquid
containing gas bubbles, USAEC Report COO-248.
Lahey, R. T. and Drew, D. A., 1992, On the development of
multidimensional two-fluid models for vapor-l i qui d two-
phase flows. Chem. Engng Comm. 118, 125-139.
Lahey, R. T., Drew, D. A., Kalkach-Navarro, S. and Park,
J. W., 1992, The relationship between void waves and flow
regime transition, Paper presented at the Int ernat i onal
Conference on Multiphase Flow Instabilities, INSA de
Roueu, France.
Lhuillier, D., 1985, Phenomenology of inertia effects in a dis-
persed solid-fluid mixture. I nt . J. Mu l t i p h a s e Fl o w 11,
427-444.
Logan, J. D., 1977, I n v a r i a n t Var i at i onal Pri nci pl es. Aca-
demic Press, New York.
Lyczkowski, R. W., Grimesey, R. A. and Solbrig, C. W., 1978,
Pipe blowdown analyses using explicit numerical schemes.
A I C h E Symp. Ser. 74, 129-140.
Matuszkiewicz, A., Flamand, J. C. and Boure, J. A., 1987,
The bubble-slug flow pat t ern transition and instability of
void fraction wave. I nt . J. Mu l t i p h a s e Fl o w 13, 199-217.
Nguyen, D. L., Winter, E. R. F. and Greiner, M., 1981, Sonic
velocity in two-phase systems. I nt . J. Mu l t i p h a s e Fl ow, 7,
311-320.
Pauchon, C. and Smereka, P., 1992, Moment um interactions
in dispersed flow: an averaging and a variational ap-
proach. I nt . J. Mu l t i p h a s e Fl o w 18, 65-87.
Richardson, S. M. and Saville, G., 1991, Blowdown of pipe-
lines. Offshore Europe 91, SPE Paper 23070.
Ruggles, A. E., Lahey, R. T., Drew, D. A. and Scarton, H. A.,
1988, An investigation of the propagat i on of pressure
pert urbat i on in bubbl y air/water flows. Trans. A S M E J.
He a t Tr ans f er l l fl , 494-499.
Semenov, N. I. and Kosterin, S. I., 1964, Results of studying
the speed of sound in moving gas-liquid systems. Tep-
l oener get i ka 11, 46-51.
Sergeev, Y. A. and Wallis, G. B., 1991, Propagat i on of con-
centration/density disturbances in an inertially coupled
two-phase dispersion. I nt . J. Mu l t i p h a s e Fl ow 17, 697-703.
Smereka, P. and Milton, G. W., 1991, Bubbly flow and its
relation of conduction in composites. J. Fl ui d Me c h. 233,
65-81.
Solbrig, C. W., Mortensen, G. A. and Lyczkowski, R. W.,
1976, An unequal phase velocity, unequal phase temper-
ature theory applied to two-phase blowdown from a hori-
zontal pipe. Pr oceedi ngs o f t he 1976 He a t Tr ans f er and
Fl ui d Me c h a n i c s I ns t i t ut e , Davis, CA.
Sursock, J.-P., 1982, Causality violation of complex-charac-
teristic two-phase flow equations. I nt . J. Mu l t i p h a s e Fl o w
8, 291-295.
Sylvester-Evans, R., 1991, Backgr ound- - t o Piper al pha tra-
gedy. Trans. I ns t . chem. engr 69B, 3-8.
Taitel, Y. and Dukler, A. E., 1976, A model for predicting
flow regime transitions in horizontal and near horizontal
gas-liquid flow. A . I . C h . E . J . 22, 47-55.
Vreenegoor, A. J. N., 1990, Macroscopic theory of two-phase
bubbl y flow. Ph.D. Thesis, Delft University of Techno-
logy.
Wallis, G. B., 1969, One- Di mens i onal Two - Ph a s e Fl ow.
McGraw-Hill, New York.
Wallis, G. B., 1988, Some tests of two-fluid models for two-
phase flow. Pr oc e e di ngs o f t he J a p a n - U. S . Se mi nar on
Two - Ph a s e Fl o w Dynami cs , Ohtsu, Japan, July 15-20.
Wallis, G. B., 1990a, Inertial coupling in two-phase flow:
macroscopic properties of suspensions in an inviscid fluid.
Mu l t i p h a s e Sci. Technol . 5, 239-361.
Wallis, G. B., 1990b, On Geurst ' s equations of inertial coup-
ling, in T wo Ph a s e Fl o ws and Wa v e s (Edited by D. D.
Joseph and D. G. Schaeffer). Springer, New York.
Yadigaroglu, G. and Lahey, R. T., 1976, On the various form
of conservation equations in the two-phase flow. I nt . J.
Mu l t i p h a s e Fl o w 2, 477-494.
APPENDIX
When the equations of mot i on are derivable from a varia-
tional principle, a general and systematic way for establish-
ing the conservative theorems can be developed from a direct
study of the invariance of the variational integral. This is
called Noether' s i nvari ant theorem (Logan, 1977; Geurst,
1985a, b).
Consider a Lagrangian L = L ( ~ , c ~ / c ~ t ~) t hat depends
only on the argument functions @i and their first derivatives
of the independent variables t ~, i = 1 . . . . . M, j = 1 . . . . . N.
The R-parameter family of transformations on the variables
t ~ and tP j can be written as
r i = dpi ( t i , ~J, g) , t~ j = goJ( t i , ~J, g) , r = 1 . . . . . R (A1)
where g denote a set of R independent parameters. The zero
value of g gives the identity transformation
i - ' = $' ( t ' , $J, 0) = t i, (~J = tpJ(ti, $J, O) = $J. (A2)
Equat i on (A1) can be expanded about g = 0 to obt ai n
~i = t i + ~ ( t i , ~ j ) s r "4- 0 ( 8 " ) , ~ J = ~ J + ~ J ( t i , l[lJ)F, " + O ( s r )
(A3)
where the Jacobi ans or the generators of the transformations
are given by
9 i a ~ J i "
~(t i , ~d) = ~, ' ( t i , J , 0) , ( i ( t i , ~ ' ) = ~7 ( t , g , ' , 0 ) . (A4)
For simplicity, the convent i on t hat repeated indices repres-
ent summat i on over the index is used.
Noether' s i nvari ant theorem states t hat a necessary condi-
tion for the fundamental integral S L d t 1, . . . , d t u to be abso-
lutely i nvari ant under the R-parameter family of transforma-
tions (A1) is t hat the following R identities hold true:
~ k - - + - - J
(A5)
where g~ is the Kronecker delta function and ~f, = O~J/63t i.
Assume t hat the fundamental integral S L d f , . . . , dt u is
i nvari ant under the M-paramet er transformation
~i=t,+g, (i,r=l... M), ~ J = ~ b j (A6)
which represents a translation of (t 1, . . . , tU)-space. The
Jacobians of this family are given by
~ = , ~ , ~ = o , ( i , r = 1 . . . . . N). (A7)
In the present work, there are only two independent vari-
ables, i.e time and space, and t I = t and t 2 = z. Equat i on (A5)
becomes, for r = 1,
Modelling of two-phase blowdown from pipelines
OIL sOLq 01" .OL'~
- + , ~ J - o ~ t + ; o + ~ ) : o
o ~L 0; ~+--T J + ~1_ ~ - 0' =~ J : o .
and for r = 2,
(A8)
(A9)
The Lag~ngi an is chosen to be the equivalent Lagrangian
density of L obt ai ned from integration by parts of, e.g., eq.
713
(18) for a one-component system:
- - S ,(~+.L ~)~- - S 2(~+u G ~s)~. (AiO)
The argument functions are given by ~, pG, PL, S, , $2, u~, uL,
~o and r/. Equation (A8) becomes the energy conservation
equation of mixtures, eq. (36), while eq. (Ag) becomes the
moment um conservation equat i on of mixtures, eq. (29).
CES 50 : 4- K
P e r g a m o n Chemical Enoineerimj Science, Vol. 50. No. 13, pp. 2173-2187, 1995
Copyright ,~(~;1995 Elsevier Science Ltd
Pnnted in Great Britain. All rights t,eservod
0009-2509/95 $9.50 + 0.00
0 0 0 9 - 2 5 0 9 / 9 5 ) t ) 0 0 0 9 - 7
MODELLI NG OF T WO- PHASE BLOWDOWN F R OM
PI PELI NES- - I I . A SI MPLI F I ED NUMER I CAL MET HOD F OR
MULT I - COMPONENT MI XT UR ES
J. R. CHEN ~, S. M. RI CHARDSON and G. SAVI LLE
Department of Chemical Engineering, Imperial College, London SW7 2BY, U.K.
(Received 11 October 1993; accepted in revised form 20 December 1994)
Abstract--A simplified numerical method is proposed to solve general two-phase flow equations for
multi-component mixtures. The method is applied to solve the marginal stability model proposed in the
first part of this paper. Case studies are performed and validated against experimental data for the
blowdown of pipelines containing one- or two-component mixtures. The results show that the marginal
stability model performs better than the simple homogeneous model for blowdown from short pipes. For
blowdown from long pipes, the results of both models are quite similar. Concentration stratification is
found to be insignificant in the overall blowdown predictions.
I. INTRODUCTION
I n the first part of this paper [ Chen et al. (1995),
hereafter referred to as Part I], a margi nal stability
two-phase flow model is proposed which possesses
real and clearly defined sonic characteristics. Critical
flow or choking, the most difficult aspect of the mod-
elling of depressurisation processes, can therefore be
studied systematically t hrough the consi derat i on of
characteristics. I n the second part of this paper, the
simplified numeri cal met hod of Chen et al. (1993) is
extended and applied to the margi nal stability model
for mul t i -component mixtures. I n particular, a de-
coupled met hod is developed to solve the conserva-
tion equat i ons of each species efficiently. The effect of
concent rat i on stratification, the change of overall
composi t i on of the two-phase mixtures due to slip
between the two phases, duri ng the rapid de-
pressurisation process is also studied.
2 . T I l E S I M P L I F I E D N U M E R I C A L M E T H O D
2.1. Generalised equations of t wo-phase phase f l ow
The generalised equat i ons of two-phase flow in
t hermodynami c equi l i bri um can be written as follows.
Mass balance of mixtures:
0
=-[ ~p~ + (1 -- ~)PL]
(Tt
1 O
+ a ~z A[~PclUG + (l -- ot)pC, uL] = 0. (1)
~'Author to whom correspondence should be addressed.
Present address: Industrial Technology Research Institute,
Center of Industrial Safety and Health Technology, Bldg. 32,
195 Chung Hsing Rd., Sect. 4, Chutung, Hsinchu, Taiwan.
Moment um balance of liquid and vapour phases:
- I LuL AI 1- +tl
= rwL + ~ + ~- (1 - ~)pLgCOsO. (2)
? I OA~ p ~ u ~ + Op
= rw~ - ri - r~ - ctp~ 9 cos 0. (3)
Energy balance of mixtures:
?P
~ [~tpGht; + (1 - ~t)pLht] dt
1 t?
+ ~ ~ A [otp~uc, htl + (! - ~t)ptuLhL]
- [ ~ u ~ + ( 1 - ~ ) u L ] Op
~- ~ - - UL' ~wL - - WG' ( wG
OZ
+ (u~ - UL)~i + Qw. (4)
Mass balance of component i:
~~(Pl Yt.i -t- P2.Vri) + ~ ~ A( pl ULyLi + P2uaYGi) = 0,
i = 1 . . . . . N - 1. (5)
where ~: denotes two-phase interaction terms not
considered in the algebraic drag z~ and p = p~,
hL = h[. zw~ is the wall drag of phase k and Qw is the
external heat source term. For completeness, the area
variation of the flow channel is also included. When
~ cont ai ns derivatives, the differential equat i ons and
their difference approxi mat i ons can be modified ac-
cordingly without much difficulty.
2.2. Finite difference method of one-component systems
The finite difference met hod used is semi-implicit
and is similar to the I CE scheme of Hariow and
2173
2174
Amsden (1971) for compressible flow. The scheme for
the homogeneous equilibrium model ( HEM) is based
on the f ollowing features (Chen e t al . , 1993):
The mixture mass equat i on is solved in conserva-
tive form to maintain mass conservation.
The moment u m convect i on term can be treated
explicitly without affecting the stability of the
scheme accordi ng to the von Neumann stability
analysis of Chen (1993).
All the other terms including the mass convec-
tion, energy convect i on and pressure gradient
terms must be treated implicitly to remove the
Courant number restriction on the integration
time-step.
The dependent variables are pressure and mix-
ture enthalpy; therefore no special t reat ment is
required for transition from single to two-phase
flows.
We now extend the simplified met hod to the general-
ised two-fluid model. Consider the simplest case of
a one-component system. It is more convenient to
define the f ollowing mixture properties based on the
mass centre average of the two phases:
p , , , = c~pG + ( 1 - : t ) p L (6)
g p G U 6 + (1 - - O : ) p L U L
u m- (7)
P,.
~tpc, h . + (1 - ~ ) p L h L
hm = (8)
Pm
Equat i ons (1) and (4) become:
~ - Ap m um= 0 (9)
~P"+~,.,z
?.t
- [ ~ u ~ + ( 1 - ~ ) u L ]
= Q~ - u , ~ w 6 - ULr~L + (U~ - UL)Z~. (10)
The moment u m equat i ons for each phase are added
to give the mixture moment u m equation:
(; 1 ~ Op
dt p " u " + A ~z A [ ~ p e u ~ . + (1 - =) pLU~] + ,3S
= Z w ~ + r , . L - - p , . O c o s O . (11)
When u~ = UL, the above equat i ons reduce to the
simple HEM. HEM is simpler to solve than a two-
fluid model because the convect i on terms can be
readily expressed in terms of mixture variables. For
two-fluid models, some special t reat ment is required.
Since moment u m convect i on can be treated explicitly,
only the energy convect i on term requires special treat-
ment, as shown below. The energy convect i on terms
are written as
~ t p e u e h 6 + (1 -- ~t )pLULh L -= pmu, , hm
+ ~tpe(l -- a ) p L (Ue - - UL}( he - - hL) . (12)
P, ,
J. R. CHEN et al.
The mixture energy equat i on becomes
?' h 1 0
d t p " " + A -~z A p , . u . , h , .
l t7 [ a p ~ ( l - - a t ) P L ( u o u L ) ( h t ~ - h L ) ]
+ ~ N A
- ~--P - [ ~ u ~ + ( I - ~ ) u ~ ] ~--P
Ot ?~z
= Qw - u ~ , - UL r , L + ( u~ - - UL ) q . (13)
The second and third terms of LHS of eq. (13) are
called the h o mo g e n e o u s and m i x i n 9 energy convect i on
terms, respectively.
Equat i ons (9), (11) and (13) are the governi ng equa-
tions for m i x t u r e properties. They are discretised
similar to. the scheme of HEM (Chen e t al . , 1993)
except for the m i x i n g energy convect i on term:
mixture mass equat i on at mass cell centre z /
V j n+l n n+l
~ t ( p , , j - p . j ) + A j + t/2(p,.u,,b+ i / 2
- A t - l:2(pmu,~)~+-~/2 = 0 (14)
mixture moment u m equat i on at moment u m cell
centre zi+ 1/ 2:
V j + 1/ 2 n - I n
At [(p,.u,,)j + 1/2 - - ( P m U m ) j + 1/ 2]
+ A ~ + l [ ~ p 6 u ~ + (1 2 .
- - 9 0 p L U L ] j + 1
n* l n+l
+ A~+x,z(pj+l - Pj )
gV~+ .+1/2 (15)
- - l l 2Pmj + 1/ 2 COS 0 1 + 1t2
mixture energy equat i on at mass cell centre z ~ :
Vj [ (p, . h. ) 7+ 1
At - (pmhm)j]
+ A j + t / 2 ( p , n u m h m ) ~ . + ~ i 2
- Aj -l / 2(pmu, ~hm)~' +-~/ 2 + At + 1:2
X ~ t P o ( ~ ) P L ( u t ~ - - UL)(/,It; - - h=)j,+,/2
"] n+ 1/ 2
x(u~ - u D ( h ~ - hL) J ~ _ 1/2
v j [ ~ . 1 _ p ) ' - I - A j r ~ u ~ + ( 1 - , ~ ) u L ] ;
A t
n+! n+l
x (pj+ 1:2 - P;- 1:2)
= E[ Qw - u~T,,~ - UL T w L
-~n+ 1/ 2
-t- ( U G - - U L J Z i J J (16)
Modelling of two-phase blowdown from pipelines--ll
where the half time-step properties, such as ( p , . ) ~ + t / 2 , + , . + t / 2
PmJ - - Pmj
etc., denote the iteration values of the n + 1 new
time-step flow variables with the old time-step values
as initial values. The n + i new time-step variables in where
the difference equations are considered as the new
iteration values. An additional outer iteration loop is
used in each time-step to ensure the convergence of
the new time-step value. All the source terms in the
mixture equations and the mixing energy convection
term involve individual phase properties and dis-
cretised using such half-time-step iteration properties
to avoid the difficulty of evaluating the individual
phase properties. An outer iteration loop is used to
achieve the convergence of these iterative properties.
Therefore, the discretised equations are the same as
HEM except that an extra mixing convection term
evaluated at the half time-step appears in the energy
equation and an outer iteration loop is also required.
The phase velocities are solved individually from the
discretised momentum equations of the two phases
with the new time-step pressure obtained from the
above mixture equations. The introduction of an
outer iteration loop certainly increases the computa-
tional load. But, since the choking or critical flow
boundary condition usually requires iterations as
well, this iterative semi-implicit method actually be-
comes advantageous over fully implicit methods in
terms of simplicity and efficiency for the present blow-
down problems. Note that the cell length, inclination
and volume are defined at the mass cell centre while
the cell cross-sectional area is defined at the edge of
mass cell. Therefore
1
V j = ~ ( A ~ + , , 2 + A s - I : 2 ) A z j (17)
Figure I shows the basic definitions of the flow chan-
nel.
The mass flux .+ I ( P , ~ Um) i * ,:2 in eq. (15) can be ex-
pressed in terms of pressure owing to the explicitly
treated momentum convection term. Therefore, the
mass flux in eqs (14) and (16) can be recast in terms of
pressure. Choosing the mixture enthalpy and pressure
as the dependent variables, the new time-step densities
can be eliminated by the following linearised equation
of state for the two-phase mixture:
g
Fig. 1. Definitions of flow channel.
- 2. n+ 1/2. n+, n+, / 2x
= ~ap,, ) j { p j - p j !
I - 2 x n + l / 2 { h n + ' n + , / 2 )
+ ~ahm ) j w, mj - - hrnj
2175
(18)
a , , = \ a p - - ~ ) , . ' a i m = \ t ~ , p = ] j (19)
Equations (14)-(16) therefore reduce to a set of two
non-linear algebraic equations. The nonlinearity only
arises from the product of two mixture variables and
can be eliminated by using the following linearisation:
f f.+,..2 \
. f " ~ ' O " ' l = , * t , ' 2 + ~ { a ; )
n* 1.'2
X (.qn+ 1,2 ~r" c q A~)
= f . ~ , . q. + , : 2 + f . + , : 2 . q . * 1
_ f . + ,;2On+ l/2 + O( A~2) . (20)
Note that the linearisation must be made along the
iteration-step not the time-step to ensure that the
higher-order truncation terms are negligible,
Finally, the linearised difference equations may be
written for cells 2 to M - 1 and solved simultaneously
with the boundary cells, cells 1 and M. to obtain the
pressure and mixture enthalpy for each cell. Details of
the difference equations are given in Chen (1993). The
solution matrix has a block tri-diagonal structure and
typical algorithms such as the Thomas algorithm can
be used to solve the algebraic equations efficiently. An
equilibrium pressure-enthalpy flash calculation is
then performed for each cell to obtain the void frac-
tion. phase densities, phase enthalpies and temper-
ature. The new-time step velocities are obtained from
the discretised momentum equation for each phase
when the new time-step pressure is known.
2.3. F i n i t e d i f f e r e n c e m e t h o d o f m u l t i - c o m p o n e n t s y s -
t e m s
The overall mass fraction of component i can be
defined as
~ PGY t ; i + (1 -- : t ) P t . Y L i
Y . i = ( 21 )
P=
Similar to eq. (12), the convection term in the mixture
species equation may also be separated into a homo-
geneous term and a mixing term:
~tpc;u~; Yc, i + (1 -- ~: ) pL u t . Y t . i = f l mu mY mi
+ ap~(l - ~ t ) P Z ( u ~ _ UL ) ( y ~ i - - Y t . i ) . (22)
X ) m
Therefore the mixture species equation becomes
-~ p.y., + ~ ~ Ap.u,.y.,
+ -A ~ z t_ p .
(uc; - UL) ( YG, - - Yl.i) 1 = O.
(23)
2176 J.R. CHEN et al.
B y using eq. (9), the above equation can be recast into
the following non-conservative form:
d c~
+ ~ A ( u G - - UL) ( YGi - - YLi ) = O.
(24)
The above equation shows that the change of com-
position is totally dependent on three factors, namely,
convective transport characterised by the mean con-
vective velocity urn, the slip velocity uG - ur, and the
volatility of the fluids characterised by the mass frac-
tion difference between the two phases. The first factor
will offset the separation of components resulting
from the latter two factors. It is therefore expected
that the maximum separation occurs in countercur-
rent flow where the mean flow is small, e.g. the case of
a distillation column. In cocurrent pipe flow, separ-
ation is expected to be diminished due to the large
mean flow. The exact extent of separation will be
dependent on the slip and volatility of the fluids.
When the two phases move at the same velocity, eq.
(24) reduces to
d
~t y' ~ + u,, -;--ez Y, i = 0 (25)
a + 1 n + 1/2
P,.j -- Pmj
which shows that concentration stratification will not
occur provided that the initial concentration is uni-
form throughout the flow channel.
Solving eq. (23) exactly by the methodology of
the previous section is very tedious and time consum-
ing since the linearised EOS becomes
, - 2~n+1/ 2. n+l n+l / 21
= l a ~ )~ [ p ~ - - p ~
n+ 1,/2)
+ (a~217 + ' / 2 ( h : + j ' - - h, ~
N- I
l - 2 - n + 1 / 2 , n+l n+l / 2~
-~ E " a~mi ) j ~Ymi j --Ymij 1
i = 1
where
(26)
a,mi \ ~P,. / . , h. (27)
and therefore dependent variables must increase as
the number of components increases. This results in
a larger coupled system and therefore larger computa-
tion time. The variable number of components also
makes the programming more difficult. Therefore,
a decoupled method is developed which solves eq. (23)
independently from other equations.
A completely decoupled species equation will im-
pose a convective Courant condition on the integra-
tion time-step as shown by Chen (1993). Since the
scheme proposed in the previous section involves an
outer iteration loop, the following decoupled dis-
cretisation for eq. (24) is used in which the iterative
half time-step is used in the convective terms instead
of the old time-step:
n-# l n A t u n + 1 1 2 , n+l / 2 n+l i 2
Ymi j = Yml j - - A Z ral l Ymi j - - Ymi j 11
A t { [ ~ t p o ( l - - o t ) p L
A j 4 . 1 / 2
Vjp~,~ pm
~n ~- 1/2
X (UG -- U L ) ( Y G i - - Y L i ) [
-l j + 1/2
- A j _ i/2 (u~ - UL)
In + 1/2
X (Y G i--Y L i)J j_ l/2 }. (28)
The above equation can be solved explicitly inside the
iteration loop for each component. It has been tested
and works well provided that the Courant number,
defined by Um, , At / Az where u r n , , is the maximum
convection speed, is not much greater than one. This
decoupled treatment therefore significantly reduces
the complexity and difficulty associated with multi-
component mixtures. The results of concentration
stratification will be presented in the next section.
2.4. Appl i cat i on to t he margi nal st abi l i t y model
Applying the simplified finite difference method to
the marginal stability model of Part I is not a difficult
task but caution must be taken to avoid possible
difficulties in the iteration. The mixture momentum
equation can be written as
l t~
~t ( p m u ' ) + A ~z A[~tpau~ + ( 1 - oQpLu 2
+ (1 - a)PLm(:t )(uG -- UL) 2 ]
&p I
+ ~ + ~ ~ [p,.(l - z )~m'(~)(u ~ - uL) 2 ]
= zw~ + zwt, - ,ap,. cos 0 (29)
where mixture variables are used where applicable
and p -- PG- Comparing with eq. {11), one notes that
two extra terms exist: the velocity fluctuations in the
liquid phase and the interfacial pressure difference.
Both terms contain a Reynolds stress coefficient func-
tion or its derivative which are given by the following
marginal stability condition:
m ( ~ ) = ~ l ' ~ - ( r h + 2 ) ~ )
I
m'(ct) = ~m - (rh + 2)or (30)
m " ( ~ ) = - ( ~ + 2) .
Substituting this equation into eq. (29) gives
c ) 1 c ~ c g p
a( ~ - (~ + 21~t) d
"4- A [ ( l - - Ot),OL(U G - - U L ) 2 ]
2A az
1 ] a ~
+ ~- ( ~+21~t (I -- 0t)PL(U~ -- UL) 2
+ ~ - (~ + 2)a
x [(1 - ~)2p~(u~ - uD ~ ]
1 2 O ~
2( ~ + 2)(1 - ct)2pL(uo -- UL) ~,Z
= z,,~ + r ~ L - - g p . , C O S 0. (31)
Therefore, under the marginal stability condition,
every term containing the derivatives of the Reynolds
stress coefficient becomes two separate terms contain-
ing the inertial coupling constant, ~ which is given by
eq. (90) of Part I. Equation (31) is the final equation to
be discretised. The first three terms of equation (31)
are the same as the HEM diseretisation. The fourth
and fifth terms originate from the flux of velocity
fluctuations and can be treated explicitly. The sixth
and seventh terms originate from the inteffacial pres-
sure difference and can be treated semi-implicitly, i..
by using the half time-step variables and iterations.
The finite difference discretisation can then be written
down without much difficulty.
Next, consider the momentum equations of indi-
vidual phase. Both equations have been used and it is
found that the gas phase momentum equation is
easier to solve both in terms of numerical stability and
complexity than the liquid phase one. Therefore, only
the finite difference discretisation of the gas mo-
mentum equation is discussed here. The gas phase
momentum equation, eq. (25b) of Part I, shows that
there are two time-dependent terms with one coming
from the inteffacial force term M~. Substituting the
marginal stability condition into M~ gives:
1 0
M, ~ = ~ , ( r ~ - ( ~ + 2 ) ~ ) ~ [ ( 1 - ~ ) p ~ ( u ~ - u , ) ]
],o
+ : t ( r h - ( ~ + 2 ) ~ t ) A Oz
x A[(1 - ~t)pL(U 6 - - UL)UG]
+ rh - (~ + 2)~t (1 - ~)p~(uo - u~)uo Oz
+~ O~ [ r f i - - ( t f i +2 ) ~ ] ( I - - ~ ) p L ( U~ - - UL ) ~ Z .
(32)
Note that the derivative in the first term contains
unknown uo and UL which can be further reduced to
one unknown only by using the following identity:
(I - ~)pL(u~ - UL)
= ( ! - = ) p ~ . I u ~ p . u . - ( 1 - ~ e)m .~ P U G q_J
= p . u a - pmu. (33)
Modelling of two-phase blowdown from pipelines--II 2177
where u,. is known since the mixture momentum
equation is solved before the individual momentum
equation. All other state variables are also known
variables. Therefore the final momentum equation of
the gas phase under the marginal stability condition
becomes
~( ~ - ( ~ + 2 ) ~ )
0
x [ ~( p. u G) - ~( p. u , . ) ] + 1 0 2 Op
~z A["PGuo] + ~
a ( ~ - (~ + 2)a) a
+ 2A ~ A[(I - ~)pL(uG -- UL)Uo]
+ - ( ~ + 2 ) ~ ( 1 - e)OL(UG -- UL )~
+ ~, [ * h - (*h + 2) a ] ( l - ~' ) # L( UG- - UL) ~
+ FUG = ~,o -- q. (34)
The discretisation is as follows: other than mo-
mentum convection, which is treated explicitly, all the
other terms can be evaluated using new iterative
values for u,. and all the state variables, and old
iterative values for uG and uL. The discretised equa-
tion can be solved to give a new ue and subsequently
a new uL can also be obtained by using the definition
of u. . Again, this lengthy difference equation can be
written down without difficulty.
Now, consider the energy equation, eq. (47) of
Part I. The derivative structure of this equation is
exactly the same as eq. (4) of the generalised model
and requires no special treatment. The final things to
be given in detail are the various source terms in the
momentum and the energy equations. In this work, all
the source terms are discretised semi-implicitly. For
example, the equilibrium entropy source term, eq. (48)
of Part I, is determined by
1
,, . + , n { [ ( 1 ~ ~ , , . . , . + . 2
= - - u ~ I P L S L j j
( r , ) s At
- [ ( l - ~ ) p L s ? . ] ~ }
1
V / { A j + 1 / z [ ( 1 . . . . . ~/2
- - ~] PLSLULJJ+ I/2
n + l / 2
- A ~ _ ~ n [ ( I - ~) P L S L UL ] s - ~ 1 2 } " ( 3 5 )
In a similar manner, the interfacial mass transfer rate
is discretised by
n+ 1 n+1/2
1 r ~ / ~ n + 1 / 2
+ ~ LAIj+ I / 2~BoUG) / + I/2
, , n+ 1/2- I
- - A j - 1/2[OZpGUG J . I - 112 J
(36)
2178 J. R. CHEN et al.
while the F, - F,,sG term in the source term FM~ is
determined by the following discretised equation:
( r , . . . . , , 2 1 . . .
- l , . s c , ~ j = ~ ( c w c , b ( s ~ ' ~ ' ' ~ - s ~ , )
1
n n~l n+l
+ ~ z J ~ p c , U ~ ) i ( s c , , . , , - s o , . . . . ) .
i
( 3 7 )
Since it is usually tedious to determine the chemical
potential difference in the multi-component mixture,
the interracial momentum flux term, FM~, is deter-
mined by the simpler equation of a one-component
system, i.e. eqs (25d) and (25e) of Part I. Note that this
simple treatment is also compatible with our de-
coupled approach in solving the mass equations for
each species where the system is solved as a frozen
mixture without composition changes within each
iteration step. If however, a fully coupled approach is
used for the species equations, then the multi-com-
ponent version of FMG, eqs (35a) and (35b) of Part I,
should be used. For the present purpose of studying
transient blowdown processes for risk assessment, the
decoupled approach is considered as sufficient and no
attempt is made to pursue a fully-coupled approach.
As we will show in the next section, the effect of
concentration stratification on the whole biowdown
process is usually insignificant.
Finally, we consider the critical flow boundary con-
dition for the marginal stability model. Once u~ and
UL are obtained, the characteristics or eigenvalues of
the model at the open end are determined numer-
ically. There are four eigenvalues [see eq. (89) of
Part I]. Two of them are the gas phase velocity and
must be positive. The other two values are the sonic
characteristics with one being negative (the left-run-
ning) and another positive (the right-running) for
sub-critical flow. Critical flow is expected to occur
when the left-running sonic characteristic equals or
exceeds zero, i.e. the downstream disturbance can no
longer propagate upstream. In this case, the velocity
of the two phases are reduced in accordance with the
magnitude of the left-running sonic characteristic. It-
eration on the half-time step variables is repeated
until the left-running sonic characteristic converges to
zero.
3. RESULTS AND DISCUSSION
3.1. H y d r o d y n a m i c c o n s t i t u t i v e r e l a t i o n s
The hydrodynamic constitutive relations are the
relations that describe the momentum interactions or
exchanges at the f luid-f luid and fluid-wall interfaces.
The momentum interactions are usually expressed in
terms of drag forces. The generalised drag force at the
fluid-fluid interfaces is usually modelled as a linear
combination of three forces: steady viscous drag, tran-
sient viscous drag or the Basset force and transient
non-viscous drag such as the virtual mass force. By
taking into account the back-flow explicitly in the
variational formulation of two-phase flow, we have
derived an inviscid two-fluid model in Part I which
includes the effect of transient non-viscous drag. On
the other hand, the transient viscous drag, usually
appearing as a time integral (lshii and Mishima, 1984),
is very difficult to incorporate as all the history of the
flow must be stored and is generally neglected. Liang
and Michaelides (1992) have studied the magnitude of
unsteady viscous drag for small bubbles for finite
Reynolds number flow. They found that unsteady
viscous drag term accounts for approximately 25% of
the total drag for bubble size smaller than 10-4m.
They also found that the unsteady viscous drag de-
creases rapidly for bubbles of larger size and larger
Reynolds number. This work will consider only the
steady part of interfacial viscous drag and neglect all
the contribution from the unsteady part.
At the fluid -wall interface, the viscous drag force is
also modelled as a linear combination of unsteady
and steady wall drag or friction. The unsteady wall
friction, usually also modelled as a time integral in
single phase flow (Zielke, 1968), can become dominant
in laminar flow but diminishes quickly with time and
in turbulent flow. Chen (1993) performed an approx-
imate study of the unsteady wall friction using single
phase expression of Zielke (1968) with the single phase
density and viscosity replaced by the two-phase mix-
ture density and viscosity of Beattie and Whalley
(1982) and found that the unsteady wall friction is
negligible compared with the steady wall friction un-
der critical flow conditions. Therefore, the transient
blowdown process will be studied by using transient
inviscid one-dimensional models with viscous bound-
ary layers at the fluid--fluid and fluid-wall interfaces
described by algebraic drag correlations to represent
viscous effects.
The interfacial viscous drag is flow-regime depend-
ent and the following simple flow-pattern map is used:
~ ~< 0.3,
0.3 < cc ~< 0.8,
~ > 0.8,
dispersed-bubble flow
intermittent flow
annular-dispersed flow.
(381
Although more mechanistic descriptions of the flow
regimes exist, e.g. the Tai t el -Dukl er (1976) flow-pat-
tern map, the large uncertainty associated with the
drag correlations usually prevents any gain in accu-
racy from using these flow-pattern maps. This simple
flow-pattern map is also widely used in modelling
steady state critical flow, e.g. Richter (1983), Dobran
(1987), and Schwellnus and Shoukri (1991).
A large amount of literature exists for the hy-
drodynamic constitutive relations. The following se-
lection are based on suggestions from the sensitivity
tests of Chen (1993). The generalised interfacial vis-
cous drag for any flow pattern is written as follows:
z i = 2 - ~ i ~ p ~ u , l u , I (39)
where u, is the slip velocity and Dh is hydraulic dia-
meter of the flow channel. The interfacial drag coeffi-
Modelling of two-phase blowdown from pipelines--ll
cient is flow pattern dependent and is given as follows:
3 D, PG
C s . i = - ~ C o - ~ r b - ~ L for 0 < ot ~< 0.3 (40)
C I i = {0.005[1 + 75(1 - ~t)](l -
E ) c t - l / 2
3 Dh ! - c t E }
+
t
for 0.8 ~< e < I (41)
Co, the drag coefficient of distorted particles (bubbles
and droplets), is given by (lshii and Zuber, 1979)
= 4 F O ( P L - - {.1 + 17.76[f(ot)]6/7~ 2
C o 5 g , [ a P~) ] 1/2 ~ j
where
f(ct) = (l - ~)L5 for bubbles, r o = rb
(42)
f(ct) = ct 3 for droplets, rp = rd.
The entrainment fraction E in annular-dispersed flow
is given by Ishii and Misbima (1984):
E = tanh(7.25 x 10- 7 We,.2s R e O . 2 5 )
w e = P ~ t 2 u 2 D h ( p L - - P G Y / 3 (43)
(1 - =)uLpLD,
R e t . =
In intermittent flow, the value of C[ ~ varies by several
orders of magnitude and is interpolated exponentially
with void fraction according to Schwellnus and
Shoukri (1991):
C[~ = x~exp(z2~) for 0.3 < a < 0.8
z 2 = l n ( C f " 2 ~ - 1- ( 4 4 )
\ Csi . l / ct~ ctt
~ = exp (In CI ~ ' ~ - ;~2 t ~ )
where the subscripts 1 and 2 denote the values at void
fraction of 0.3 and 0.8, respectively.
The inteffacial viscous drag discussed above deter-
mines the slip between the two phases at a continuum
point in a general three-dimensional flow field. In
bubbly or droplet flow, such local slip is actually
relatively small compared to the mean flow velocity.
In one-dimensional flow, the non-uniform distribu-
tion of the two phases in the transverse direction of
the flow channel also affects the overall one-dimen-
sional slip between the two phases. The effect of phase
distribution still exists even when the local slip is zero
and is usually larger than the local slip (Ishii and
Mishima, 1984). Therefore, only the averaged local
relative velocity rather than the difference between
averaged velocities should be used in the above rela-
tions for averaged interfacial drag in one-dimensional
flow.
In bubble and intermittent flow, the averaged local
relative velocity, denoted by ti,, is related to averaged
velocities by (Ishii, 1979)
2179
I - Coot
u , ~- uG - C o u L (45)
1--~t
where Co is the distribution parameter. For bubble
flow, the following empirical correlation of Co for
a round tube is used (Ishii, 1979):
Co = 1.2 - 0.2 P f~t6. (46)
For annular or stratified flow, the distribution para-
meter is generally very close to one and the phase
distribution effect can be neglected (Ishii, 1977). For
intermittent flow, a linear interpolation between
Co = I at ~t > 0.8 and Co given by eq. (46) at x < 0.3 is
used:
= _ _ _ ~g ~< ct ~< ~t2.
\ct2 - ~/ '
(47)
T h e f r i c t i o n r e s u l t i n g f r o m t h e vi scous b o u n d a r y at
t h e wa l l f o r t w o - p h a s e f l o w is s i mi l a r t o t h e si ngl e
phase one. I t i s u s u a l l y r e p r e s e n t e d b y t h e s i ngl e phase
w a l l f r i c t i o n m u l t i p l i e d b y a t w o - p h a s e mu l t i p l i e r .
Since gas is rarely in contact with the wall, the wall
friction is assumed to arise from the liquid phase only
and
rwL 4)22 CfLo 62
= - (48)
D, PL
where the two-phase multiplier 4)20 of Friedel (1979) is
used.
Finally, the wall to fluid heat transfer rate per unit
volume, Qw, is given by
4
Qw = ~ hw(Tamb -- T6) (49)
where hw is overall heat transfer coefficient, T~mb is the
ambient temperature and the fluid temperature is
chosen to be T6.
3.2. M o d e l v a l i d a t i o n : o n e - c o m p o n e n t s y s t e m
A computer program called META (Multi-com-
ponent Equilibrium Two-phase Analyser) has been
developed based on the present simplified numerical
method and incorporating both the HEM and the
marginal stability two-fluid model. The thermodyn-
amic constitutive relations, namely the phase equilib-
rium relations and thermophysical properties of the
fluids, are provided by a computer program called
PREPROP which is a general thermodynamic pack-
age developed in the Fluids group in the Chemical
Engineering Department of Imperial College. PREP-
ROP incorporates two different methods for predic-
ting phase equilibrium and thermodynamic proper-
ties. The first method is a corresponding states
principle (CSP) based on an accurate equation of state
for methane (Saville and Szczepanski, 1982), and the
second method is the Peng-Robi nson (Peng and
Robinson, 1976) equation of state (PR). PR is gener-
ally more efficient than CSP while CSP predict more
2180 J. R. CrtEt~ e t al .
accurate mixture properties if the mixtures are all
similar to methane, Both methods has been used in
the validations.
To validate the finite difference method indepen-
dently from the interference of real fluid effect, an
extra option of perfect gas properties is built into
META. The perfect gas option of META is used to
simulate the blowdown of a 1000 m long pipe and
compared to the numerical benchmark solution of
wave-tracing method-of-characteristics (Chen e t a l . ,
1992). The comparison shows that the simplified finite
difference method is about as efficient and accurate as
the wave-tracing algorithm of method-of-character-
istics (Chen, 1993). Another validation is also made
using HEM against the numerical benchmark solu-
tion of Hancox e t a l . (1976) for the blowdown of
a subcooled water line that emulates the blowdown
experiment of Edwards and O' Brien (1970). Again, the
present numerical method agrees well with the numer-
ical benchmark as show in Chen (1993). Nevertheless,
the agreement between the predictions and the experi-
ment is not completely satisfactory; in particular, the
predicted pressure always drops more quickly than
the measurements. Since the HEM usually predicts
the lower limit of critical flow data (Ardron and Fur-
ness, 1976), the correct predicted pressure should drop
more slowly rather than more quickly than the
measurements. Solbrig e t a l . (1976) suggested that the
discrepancy arises from the uncertainty in the pipe
rupture mechanism, e.g. some obstacle remains at the
circumstance of the pipe. As the exact pipe bore size is
unknown, the experiments of Edwards and O' Brien
(1970) will not be used for model validation. Alterna-
tive blowdown experiments performed by Necmi and
Hancox (1978) are used for model validation instead.
The experiment is described briefly here. The pipe is
horizontal, 4 m in length, 0.032 m in diameter and was
made from transparent PVC. The pipe is filled with
dichlorotetrafluoroethane (Freon ! 14 or R 114) and
pressurised to 15 bar. The temperature is maintained
at room temperature around 21.4'JC. The saturation
pressure of R 114 at this temperature is about 2 bar.
One end of the pipe is closed and the other end is
fitted with a combination of a 0.0005 m thick glass
disc and a diaphragm. The glass disc and the dia-
phragm are ruptured almost simultaneously by
a plunger in less than 0.5 ms. Necmi and Hancox
(1978) reported that this arrangement gave a clear
break to the full pipe bore.
The calculations are made by assuming a constant
wall heat transfer coefficient of 5 W/m2K and a con-
stant roughness length scale of 0.0015 mm. Varying
these parameters does not have any significant effect
on the results as shown in Chen (1993). 25 uniform
meshes of 0.16 m and time-step of 0.0004- 0.04 s were
used. Further mesh refinement does not have any
significant effect on the result. The calculation is made
with the Peng-Robinson equation of state. Since R-
114 is significantly different from methane, the
methane-based CSP is not used. Each calculation
takes about 40 min on a DEC 5000/240 station. Fig-
ure 2(a)-(d) shows the results of pressure at 0.86 m
from the closed end, pressure at 0.04m from the
ruptured end, void fraction at 0.86 m from the closed
end and void fraction at 1.06 m from the ruptured
end, respectively. Both the results of HEM and the
two-fluid marginal stability model are shown.
The results of HEM show pressure dropping more
slowly than the measurements of Necmi and Hancox.
This is in line with the observation that HEM predicts
a lower critical flow rate than the actual flow rate out
of the line. The measured pressure drops quickly be-
low the saturation pressure at the point when the
rarefaction wave reaches the closed end of the line but
never recovers to the saturation pressure. It is quite
clear that thermodynamic non-equilibrium, i.e. de-
layed bubble nucleation and slower vapour genera-
tion rate, is dominant in this stage and the observed
flow pattern is bubbly flow. After 0.4 s, the pressure
starts to drop again until the end of blowdown. The
observed flow patterns less than and greater than I m
from the ruptured end are dispersed droplet flow and
entrained stratified flow, respectively. The calculated
void fraction history at 0.86 m from the closed end
shows that around 1 s the calculated void fraction is
underestimated compared with the experiments. Since
thermodynamic equilibrium ensures the maximum
vapour generation rate, this underestimation suggests
the existence and the importance of mechanical non-
equilibrium or non-homogeneity at this stage and the
outflow rate is expected to be underestimated con-
siderably by the HEM prior to 1 s of blowdown. The
observed flow patterns, void fraction and pressure
histories suggest both thermal and mechanical non-
equilibria are important at this stage.
The pressure at the closed end predicted by the
marginal stability model (MSM) remains at the satu-
ration pressure until about 1 s after which the pres-
sure drops more quickly than the measurements and
predictions of HEM. Previously observed over-es-
timation of void fraction around 1 s at 0.86 m from
the closed end for HEM is found to have better
agreement with the prediction of MSM. This confirms
the existence and importance of mechanical non-equi-
librium. The predicted pressure of MSM at the rup-
tured end drops to 1.1 bar within 0.1 s and then de-
creases slowly until the end of the blowdown. This
result is actually a combination of the assumption of
thermodynamic equilibrium and the inaccuracy of the
MSM at high void fraction. We mentioned in Part
I that the marginal stability model predicts the sonic
characteristics well up to a void fraction of about 0.8.
At a higher void fraction, the model requires a higher
flow velocity to reach the choking condition. This is
because the model still predicts negative character-
istics as shown in Fig. 3 of Part I while the real flow
should already have choked. The consequence of this
deficiency is that the predicted critical flow rate will be
higher than the actual flow rate and therefore results
in a quicker pressure drop. Thermodynamic equilib-
rium further enforces the maximum possible vapour
generation rate and brings the flow quickly into a high
_ g
"6
>
Modelling of two-phase blowdown from pipelines--II
(a) Pressure histories at 0.86 m from
the closed end.
2. 4.
2.2
2.0-
, , " . . , ,
1. 8" " x .
1.6- ".
~ ' "' 4
1 . 4 , ~ ",
1 "2 I
1.0
0 . 8 I I ~ ~ I
0 0 . 5 1 . 0 1 . 5 2.0 2.5
Time (s)
2.4
2.2.
2.0
(b) Pressure histories at 0.04 m from
the raptured end.
Measurement
. . . . . . . META-HEM
. . . . META-MSM
. . . . . . . . . . . . . . . . . ,
1.8
! . 6 ~ "',,
1. 4 ""'"'.-,.
t
~ , ".
0 . 8 t i
0 0.5 1.0 1.5 2.0
Time (s)
2.5
(c) Void fraction histories at 0.86 m from
the closed end.
| . 0 j ~-- ." ."
o 9 T /
0 " 8 / ""
/ I
I
o.7Jf
0.6 I 0.5
0.4
0.3 /
0. 2 xx x
0.1 ~//
0 0/5
/ l x" x
l x . '
i / X ."
I ..
x o.
I .' "
t I
I / 0 1.5 2. 0
Time (s)
1.0
0.9

0 . 8
0.7
0.6
0.5
~ 0.4
0.3
0.2
0.!
2.5 0
(d) Void fraction histories at 0.04 m from
the ruptured end.
I
i /
l I
I
I x
I
I
I
I x
I
I
I x
I
I
I .
o.'5
. _ _ - - - / ; .....

Measuremen
. . . . . . META-HEM
. . . . META-MSM
1 . 0 1 . 5 2 1 0
Time (s)
2.5
Fig. 2. Results of one-component systems.
2181
void fraction regime as shown in Fig. 2(d). When
thermodynamic non-equilibrium is incorporated into
MSM, it is expected that delayed vapour generation
will result in a lower void fraction near the rupture
plane and choking is expected to occur at a lower void
fraction and therefore make pressure drop more
slowly. Another reason why the predicted pressure in
this particular case seems to deviate considerably
from the measurements is that the saturation pressure
is very close to the ambient pressure. In high pressure
systems as we shall see in next validation, the magni-
tude of deviation is about the same as for this system
which is actually within 1 bar [see Fig. 4(a) and (b)].
The calculated void fraction histories at 1.06 m from
the ruptured end shown in Fig. 2(d) show that the
actual void fraction is lying between that predicted by
MSM and HEM with preference to that of MSM. The
HEM, once again, underestimates the void fraction
and this suggests the importance of mechanical non-
equilibrium.
Finally, we compare the mass flow rate predicted by
MSM and HEM in Fig. 3. The maximum mass flow
rate predicted by MSM is about 4-5 times larger than
that of HEM in the first 0.1 s. The true mass flow rate
is expected to lie between the two. However, the exact
mass flow rate in this experiment can only be accu-
rately predicted by relaxing the thermal equilibrium
assumption. Although thermodynamic non-equilib-
rium can be incorporated in this work for this particu-
lar case, it will not be significant in the problems we
are mainly interested, namely, multi-component two-
phase flow in long pipes (/> 100m). In next section,
the comparison of predictions and measurements for
blowdown from a l oom pipe will confirm this argu-
ment.
3.3. Mo d e l validation: t wo- component s y s t e m
Very few experimental or field data exist for transi-
ent multi-component two-phase flow. The most com-
prehensive are the full scale experiments performed by
. . . . . . . META-HEM
. . . . META-MSM
2182
1o
~ s
h,
6
o
4
m
2
. -. . . . . . . . . . . . . _ . ' : . - . . . . . .
- - - +_ _ - - - - - - - . . ~ ~ - - - - __q
0. 5 1.0 1.5 2. 0 2. 5
Time (s)
Fig. 3. Comparison of calculated mass flow rate histories at
the rupture plane for HEM and MSM.
British Pet rol eum and Shell Oil [see Tam and
Cowley (1988)] on the Isle of Grain, England.
The experiments are very extensive and cont ai n not
only transient measurement s inside the pipe but also
J. R. CHEN e t a l .
j et dispersion outside the pipe, Onl y the i nf ormat i on
relevant to this work is described here. The pipe is
horizontal, 100 m in length, 0.15 m in di amet er and
made from commerci al steel. The pipe is filled with
pressurised LPG cont ai ni ng 95 mol e% propane and
5 mol e% butane. The test pressure varies f rom 8 to
21 bar and the t emperat ure is ambient t emperat ure
varying between 15 and 2ff~C. The saturation pressure
of the mixture around this t emperat ure is about 8 bar.
One end of the pipe is closed and the ot her end is
fitted with a di aphragm to the desired bore size. The
bore size is controlled by a circular orifice plate and
varies f rom full-bore to 0.05 m in diameter. Tam and
Cowley did not ment i on the mechani sm used to rup-
ture the di aphragm but they reported that the rupture
is rapid and produces an unobstructed full-bore aper-
ture. Tam and Cowley also gave the roughness of the
pipe which is characterised by a length scale of
0.05 mm. Pressure and t emperat ure are measured at
ten different locations al ong the pipe. The pipe is also
weighed by 20 unif ormly spaced Ioadcells. To f urther
characterise the flow, eight neutron back-scattering
sensors are installed to measure the void fraction.
12
!
(a) Pressure histories at closed end.
~ . . . . . .
~ 6
t
o 5 , o , , 25
Ti me (s)
( c) Temperature histories at cl osed end.
3 o ]
2 o i
' I
-I0
E o -20
F-
-30
-40
-50
0
~~ ~' ~~
I 0 20 25
Time (s)
12
I0
8
(b) Pressure histories at open end.
c
t-.
Me a s u r e me n t
. . . . . . . META-HEM
. . . . MSM-CS
. . . . . MS M- n o CS
6 ' - , . . . . .
4 ' \ , "'""" ......
I
0 5 I0 15 20 25
T i me (s)
(d) Te mp e r a t u r e h i s t o r i e s at o p e n e nd.
20 L Me a s u r e me n t
l Og N. ~ . . . . . . . ME T A - H E M
]~ ~ . . . . MSM-CS
M S M - n o C S
I0 "",
0 5 10 15 20 25
Time (s)
Fig. 4(a)-(d). Results of two-component systems.
Modelling of two-phase blowdown from pipelines--II 2183
1.2
1.0
0.8
o
"~. 0. 6
g
N o.4
o
0.2
O"
-0.2
0
I000-
900-
800-
700"
o SO0-
400-
300"
2 0 0 "
I 0 0 "
(e) Voi d fracti on histories at closed end.
I . ' "
)
5 I 0 15 20 25
Time (s)
1 .0 .
0 "8 i
i
0 .6 "1 :
(f) Voi d fraction histories at open end.
~
J
. . . . MSM-CS
0, I
0.2
o
. . . . . MSM-no
) ) i
I 0 1 5 2 0 2 5
Ti me (s)
(g) Total inventory of line.
" ' ~ t ~ , Load cell
"~'.'.: .~ t . . . . . H o l d u p
~ , ~ " ~ ; . . . . . . . M E T ^ - H E M
"~"~:'. - - - - MSM-CS
~N~'~'~,j ~ . . . . . MSM-no
"x~--~: .~ .
"., "l 'v':r L
" ;.. ' ','.,.: , . ,
- ~' , , :.:.,,.,,
10 15 20 25
Time (s)
Fig. 4(e)-(g). Results of two-component systems.
The measurements chosen here for validation are
for a full-bore rupture and the initial pressure and
temperature are 11.25 bar and 19.9C, respectively.
The calculation is made by assuming a constant wall
heat transfer coefficient of I00 W/m2K and a constant
roughness length scale of 0.05 ram. 25 uniform meshes
of 4 m and time-step of 0.005-0.02 s were used. Fur-
ther mesh refinement does not have significant effect
on the result. The Peng-Robinson equation of state is
used for all calculations of thermodynamic and phase
behaviour. The computation time is about 20 h for the
marginal stability model and 8 h for the homogeneous
model of which 95% of the CPU time is spent on the
thermodynamic and phase equilibrium calculations.
Varying the above parameters does not have a signifi-
cant effect on the predictions except the roughness
length scale, which we will discuss later. More valida-
tions are given in Chen (1993).
Figure 4(a)-(g) shows the results for pressure at
closed and open ends, temperature at closed and open
ends, void fraction at the closed and open ends, and
inventory of the pipe, respectively. By closed and open
ends we mean the gauges or sensors that are closest to
the closed or open ends. For pressure and temper-
ature gauges, they are 1.5 m from closed end and
0.14 m from open end. For void fraction sensors, they
are 1.74 m from the closed end and 0.18 m from the
open end. The calculated results shown are the values
at cells closest to open and closed ends. Two different
measurements of inventory are shown. The first one is
based on the direct weighing by the load cell (legend
"load cell"). The second method is based on the liquid
holdup measurement from the neutron back-scatter-
ing (legend "holdup").
The measured pressure at the closed end shows
a rapid drop to close to the saturation pressure and
then stays around the saturation pressure until 12 s
after which the pressure drops gradually to ambient
pressure. This result shows that the effect of thermo-
dynamic non-equilibrium in long and large diameter
pipe is not so significant compared to that in small
pipes. The most significant non-equilibrium effect, the
delayed bubble nucleation shown by the pressure
undershoot below the saturation pressure, is probably
2184
dominant only in the first 0.5 s. Within 1 s the pres-
sure recovers to above the saturation line and the
remaining results do not seem to be affected by the
delayed nucleation.
The comparison of measurements and HEM pre-
dictions are surprisingly good. All the predictions of
pressure, temperature, void fraction and inventory
histories almost coincide with those of measurements.
The only exception is the void fraction at the closed
end which is over-predicted after 7 s. The HEM pre-
dicts a rapid increase of void fraction after 7 s while
the measured void fraction does not show a rapid
increase until 10 s. This discrepancy is attributed to
another effect of thermal non-equilibrium in which
the vapour generated is less than the equilibrium
value and therefore the measured increase in void
fraction is slower than the prediction. Nevertheless,
the good agreement between the measurement and
HEM predictions of pressure and temperature sug-
gests that the state of the two-phase mixture is not
far from equilibrium. One other point to note is that
the predicted pressure drops slightly more quickly
(rather than slowly) than that measured. However, as
the discrepancy is relatively small, it is difficult to
judge whether the quicker pressure drop is a result of
uncertainty in the rupture mechanism or in other
sources. Another possibility of quicker predicted pres-
sure drop could be the smaller roughness length scale.
Figure 5 shows the predicted pressure at the closed
and open ends for roughness length scales varying
from 0.0 to 0.02 mm. If the pipe is assumed to be
smooth, the pressure can be under-predicted signifi-
cantly at the closed end. We take this opportunity to
correct an argument in Chen e t al . (1993). In that
paper, it was mistakenly said that the HEM over-
predicts the discharge rate because the pipe is as-
sumed smooth. This is incorrect as we discussed be-
fore. By increasing the roughness to larger than
0.1 mm, it is possible to get the predicted closed end
pressure higher than the measurement. Nevertheless,
J. R. CHEN et al.
the pressure at the open end is still under-predicted
before I I s which is the time when pressure/void
waves reflect back to the open end. During this per-
iod, an increase of roughness will result in a decrease
of pressure at the open end as one can see from Fig. 5.
As we already showed in Fig. 2(b) that HEM should
predict a higher pressure even at the open end, we
therefore suggest that the discrepancy in the pressure
drop prediction does not result from the wall friction
but perhaps results from an obstacle that remains at
the pipe end after the pipe is ruptured. Uncertainty in
the roughness length scale is still possible but it is
considered less significant than the uncertainty of the
rupture mechanism.
Figure 4(a)-(g) also shows the predicted results of
MSM with and without concentration stratification.
As there could be some uncertainty associated with
the measurements, good agreement between MSM
and the measurements is n o t expected. The first thing
to note in these results is that the results of MSM with
and without concentration stratification (CS) are al-
most indistinguishable. This is not unexpected as we
mentioned in Section 2.3 that the mean flow velocity
will offset the change of composition due to the slip. In
the present case, the mean flow velocity is greater than
100 m/s while the slip is only about 20 m/s as shown
in Fig. 6. The change of overall composition as shown
in Fig. 7 is found to be less than 5% before choking
ends and increases thereafter to about 50% at the
closed end and 15% at the open end. The change of
composition is larger at the closed end than the open
end because the mean flow velocity for the former is
smaller than for the latter. Nevertheless, the large
change of composition does not have any significance
on the predicted pressure, temperature or inventory
histories because the inventory after choking ceases is
quite small.
Returning to the results in Fig. 4(a)-(gL both the
predicted pressure and temperature at both ends of
the pipe are lower than the HEM predictions and
12
1 0 ~ . . ~ C ! o s e d / e n d
~$~_ ~" ~. . ~. ~. , 0 . 2 m m
~.^. ~ ~ O. l mm
. o ",~.:~Smooth" \~"..'.~.
" ; . - . x "', "\ ,'
Az " ~' ; ~' ~' ~' - ' ' ~ 0.05 m m
2 Open end ""~i\x~,: i ~
, i
0 5 10 15 20 25
Time (s)
Fig. 5. Effect of roughness on predicted pressure histories.
120 t
.,,' t V - ' ~ ~.ak" apour
~' ~" . . . . " ~ 17 velocity
. ~ 8 0
> 6 0 -
40-
. . . . MSM-C$ ',', \
20- . . . . . . . MSM-no CS "~,,.~._....
0 5 10 15 20 25
Time (s)
Fig. 6. Comparison of predicted flow velocities at rupture
plane of HEM, MSM and MSM-CS.
0.075
Modelling of two-phase blowdown from pipelines--ll
250.
2185
0.070
r~
0.065
Q
.~,
o 0.060
o
E
0.055
200
I '
z"
/ ' ..'
4 m from closed end / .'
Closed end~ ./'Y'
4 m from open e n d \ ~ / : ' : \ \
Open end! ~" " ~. / . ; ~
5 10 15 20
Time (s)
150
_~ 100
50
- - META-HEM
. . . . MSM-CS
. . . . . . . MSM-no CS
0.05
0 25 0 25 5 10 15 20
Time (s)
Fig. 7. Predicted variations of total mole fraction of butane Fig. 8. Comparison of predicted release rate of HEM, MSM
in different locations of pipe. without CS and MSM with CS.
measurements while the void fraction is higher at both
ends. This suggests the out flow rate is over-predicted
by MSM. The over-predicted release rate is con-
sidered a result of the uncertainty of the experiment
and inaccuracy of MSM itself. However, the good
agreement of inventory history of Fig. 4(g) shows that
the release rate of both models might not differ signifi-
cantly. The comparison of the release rate of both
models is shown in Fig. 8. Except for the first second
of blowdown, the predicted release rate of MSM (both
with and without CS) and HEM are quite close. This
result confirms that HEM is a good approximation
for strongly coupled two-phase flow in long and large
diameter pipes.
Finally we compare the present predictions with the
other work. The Isle of Grain experiment has also
been compared with predictions from the BLOW-
DOWN code of Richardson and Saville (1991) and
the PLAC code (Hall e t a l . , 1993). The BLOWDOWN
code for pipelines is developed based on quasi-steady
state, equilibrium and homogeneous two-phase flow
assumptions. It is expected that, provided the time
step used in BLOWDOWN is small enough, the re-
sults will approach those of HEM. The thermodyn-
amic package used by BLOWDOWN is the same as
this work. PLAC (Phiibin and Govan, 1990), on the
other hand, is a shortened version of the nuclear safety
code TRAC (LANL, 1986) with a different built-in
thermodynamic package for handling the phase beha-
viour of multi-component mixtures. The conservation
equations solved by TRAC and PLAC are similar to
the ill-posed Wallis model except that thermodynamic
non-equilibrium is allowed. This poses the question of
which temperature should be chosen for the equilib-
rium phase behaviour as there is no other way except
a thermodynamic equilibrium condition to determine
the phase behaviour of multi-component mixtures.
This is the problem of allowing thermal non-equilib-
rium in the two phases for multi-component mixtures
which is noticed and avoided in this work by using the
consistent equilibrium assumption. Also, the original
critical flow boundary condition in TRAC, the Trapp
and Ransom (1982) criterion which consider the real
part of the characteristics to be zero if the character-
istics are complex, is replaced by a homogeneous
frozen flow model which is not consistent with the
flow model in PLAC.
The results are shown in Fig. 9(a) and (b) for pres-
sure and inventory histories, respectively. The predic-
tion of BLOWDOWN is found to be quite close to
that of HEM at the open end except for the initial
rapid pressure drop which the quasi-steady assump-
tion cannot resolve properly. The predicted pressure
at the open end is found to be close to the result of
MSM rather than the HEM. It is not clear why this
difference arises. The predicted inventory history is
slightly lower than the HEM but is still within the
uncertainty of the experiment. The predicted initial
inventory of the line is lower because a methane-based
CSP is used. Overall, the BLOWDOWN code does
a good j ob j udged on its simplifications. PLAC, on the
other hand, performs surprisingly poorly. Firstly, the
sharp drop of pressure at the closed end is not re-
solved but shows smearing similar to the quasi-steady
assumption of BLOWDOWN. Secondly, the pressure
at both ends drop almost simultaneously by over
2 bar about 5 s after the start of the blowdown. This
behviour is not observed in the measurements or the
present predictions and the reason is not clear. Most
of all the predicted inventory, shown in Fig. 9(b),
shows a surprisingly low initial inventory. Hall e t a l .
(1993) mentioned that the initial state of the fluid is
80% liquid. Apparently, this results from incorrect
thermodynamic predictions. Also the predicted inven-
tory drops to less than 20% of its initial value within
5 s. Clearly, the release rate is considerably over-es-
timated by the homogeneous frozen critical flow
boundary condition. In summary, the poor perfor-
mance of PLAC suggests the importance of accurate
predictions of thermodynamic and phase behaviour
2186
(a) Pressure histories at closed and open ends of line.
12, -
Measur ement
. . . . . . . META- HEM
. . . . MSM- CS
10-
. . . . . BLOWDOWN
\ - . . . . PLAC
8- ~ Closed end
,~ [ ' - - . : ~ .
, . . . . . . . . . " : b ~ . - - . ? , k
/ O p e n e n d - " . - " ~ . : . ' , . N ,
o ; ,o 15 2 0 2 5
Time (s)
(b) Total i nvent ory of line.
1900~ t 000 ~- - . . . . Load cell
" \ --. Hol dup
s 0 o ~ , - M~TA-.EM
. . w
0 0 4 - . . . . . . L A C
\
! -
\ h . k I
t
I
0
J. R. CHEN et al.
a perfect gas and the homogeneous two-phase
flow model.
In full-bore blowdown from short pipes
(< 10m), both thermal and mechanical non-
equilibrium are significant. Validation against
Necmi and Hancox' s measurements shows that
the release rate is significantly under-predicted
by the HEM but is slightly over-estimated by the
marginal stability two-fluid model which allows
mechanical non-equilibrium. To model the
blowdown behaviour more accurately, relax-
ation of thermal equilibrium assumption is sug-
gested.
In blowdown from long pipes ( I> 100 m), thermal
non-equilibrium is found to be insignificant ex-
cept for the very early stages of rarefaction wave
propagation. The results from HEM and MSM
show good agreement with the Isle of Grain
experiments. In particular, the release rates pre-
dicted by HEM and MSM are very close to each
other.
The effect of concentration stratification in the
Isle of Grain experiments is found to be insigni-
ficant. It is suggested that, in transient blowdown
processes in simple geometry pipelines contain-
ing multi-component mixtures, this effect can be
generally neglected.
Thermodynamic and phase equilibrium calcu-
lations are found to occupy at least 95% of the
CPU time of the total calculations. Accurate
predictions of the thermodynamic and phase
equilibrium behaviour are essential to transient
flow calculations. But more efficient thermodyn-
amic calculation methods are clearly desirable.
5 10 15 20 25
Time (s)
Fig. 9. Comparison of results of META, BLOWDOWN
and PLAC.
and the consistency of the two-phase model and criti-
cal flow boundary condition in transient blowdown
calculations.
4. CONCLUSI ONS
A simplified numerical method is developed for
solving general equilibrium multi-component two-
phase flow models. Extension of this method to the
proposed marginal stability model is demonstrated.
Case studies have been made with the homogeneous
equilibrium model and the marginal stability model
and these have been compared with experimental
measurements for two-phase blowdown from pipes
containing one- and two-component mixtures. The
most important conclusions are as follows.
The simplified numerical method is found to be
satisfactory for blowdown problems through
validations against the wave-tracing solution for
Acknowl edyement s Financial support from British Gas pie
for JRC through the award of research scholarship is grate-
fully acknowledged.
NOTATION
A cross-sectional area of flow channel
Co distribution parameter
Cs~ interracial drag coefficient
Dh hydraulic diameter of flow channel
E entrainment fraction
hk Specific enthalpy of phase k
h,, mixture specific enthalpy
g acceleration of gravity
G mass flux
m Reynolds stress coefficient
inertial coupling constant
M~ interfacial force density
N number of components of fluid
p pressure
Q~ external heat flux
rp radius of particles
t time
T~ fluid temperature
T~mh ambient temperature
uk velocity of phase k
u,, mixture velocity
Modelling of two-phase blowdown from pipelines--II
~, averaged local relative velocity
V~ vol ume of comput at i on cell j
yk~ mass f raction of component i in phase k
y, , ~ mixture mass f raction of component i
Y~j.j=I.2 reduced densities of liquid and vapour
phases of component i
z spatial coordi nat e
Greek l et t ers
~t void f raction
At time-step
Az space-step
I'm interfacial mass transfer rate
Fuk interfacial moment u m flux of phase k
F~ interfacial equi l i bri um ent ropy transf er rate
/~L viscosity of liquid
pj, j : l . 2 reduced densities of liquid and vapour
phases
pk density of phase k
pm mixture density
a surface tension
r~ interfacial viscous drag
rwk wall drag of phase k
REFERENCES
Ardron, K. H. and Furness, R. A., 1976, A study of the
critical flow models used in reactor blowdown analysis.
Nucl. Engnq Des. 39, 257-266.
Beattie, D. R. and Whalley, P. B., 1982, A simple two-phase
frictional pressure drop calculation method. Int. J. Multi-
phase Flow 8, 83-87.
Chen, J. R., 1993, Modelling of transient flow in pipeline
blowdown problems. Ph.D. thesis, University of London.
Chen, J. R., Richardson, S. M. and Saville, G., 1992, Numer-
ical simulation of full-bore ruptures of pipelines contain-
in~ perfect gases. Trans. Instn chem. Engrs 70B, 59-69.
Chen, J. R., Richardson, S. M. and Saville, G., 1993, A simpli-
fied numerical methbd for transient two-phase pipe flow.
Trans. Instn chem. Engrs 71A, 304-306.
Chen, J. R., Richardson, S. M. and Saville, G., 1995, Model-
ling of two-phase blowdown from pipelines--I A hyper-
bolic model based on variational principles. Chem. Enong
Sci. 50, 695-713.
Dobran, F., 1987, Nonequilibrium modeling of two-phase
critical flow in tubes. Trans. ASME J. Heat Trans. 109,
731-738.
Edward, A. R. and O'Brien, T. P., 1970, Studies of phe-
nomena connected with the depressurization of water re-
actor. J. Br. Nucl. Eneroy Soc. 9, 125-135.
Friedel, L., 1979, Improved friction pressure drop correla-
tions for horizontal and vertical two-phase flow. Proceed-
ings of the European Two-Phase Flow Group Meeting,
lspra, Italy.
Hall, A. R. W., Butcher, E. R. and Teh, C. E., 1993, Transient
simulation of two-phase hydrocarbon flows in pipelines.
2187
Proceedings of the European Two-Phase Flow Group Meet-
iny, Hannover.
Hancox, W. T., Mathers, W. G. and Kawa, D., 1978, Analysis
of transient flow-boiling: application of the method of
characteristics. A.I.Ch.E. Syrup. Ser. 74 (174), 175-183.
Harlow, F. H. and Amsden, A. A., 1971, A numerical fluid
dynamics calculation method for all flow speeds. J. Corn-
put. Phys. 8, 197-213.
Ishii, M., 1977, One-dimensional drift-flux model and consti-
tutive equations for relative motion between phases in
various two-phase flow regimes. Argonne Natl. Lab. Re-
port ANL-77-47.
lshii, M. and Mishima, K., 1984, Two-fluid model and hy-
drodynamic constitutive relations. Nucl. Engng Des. 82,
107-126.
lshii, M. and Zuber, N., 1979, Drag coefficient and relative
velocity in bubbly, droplet or particulate flows. A.I.Ch.E.
J. 25, 843-855.
LANL (Los Alamos National Lab.) , 1986, TRAC-
PFI / MODI : An advanced best estimate computer pro-
gram for pressurized water reactor thermal-hydraulic ana-
lysis. Los Alamos Natl. Lab. Report LA-10157-MS,
NUERG/CR-3858.
Liang, L. and Michaelides, E. E., 1992, The magnitude of
Basset forces in unsteady multiphase flow computations.
ASME J. Fluid Engng 114, 417-419.
Necmi, S. and Hancox, W. T., 1978, An experimental and
theoretical investigation of blowdown from a horizontal
pipe. Proceedings of the 6th International Heat Transfer
Conference, Toronto, Canada.
Peng, D. Y. and Robinson, D. B., 1976, A new two-constant
equation of state. Ind. Engng Chem. Fundam. 15, 59-64.
Philbin, M. T. and A. H. Govan, 1990, Pipeline Analysis
Code. UKAEA Report AEA-APS-0031.
Richardson, S. M. and Saville, G., 1991, Blowdown of pipe-
lines. Offshore Europe 91, SPE Paper 23070.
Richter, H. J., 1983, Separated two-phase flow model: ap-
plication to critical two-phase flow. Int. J. Multiphase
Flow 9, 511-530.
Saville, G. and Szczepansk, R., 1982, Methane-based equa-
tions of states for a corresponding states reference sub-
stance. Chem. Engng Sci. 37, 719-725.
Schwellnus, C. F. and Shoukri, M., 1991, A two-fluid model
for non-equilibrium two-phase flow. Can. J. Chem. Engn#
69, 188-197.
Solbrig, C. W., Mortensen, G. A. and Lyczkowski, R. W.,
1976, An unequal phase velocity, unequal phase temper-
ature theory applied to two-phase blowdown from a hori-
zontal pipe. Proceedings of the 1976 Heat Trans. Fluid
Mech. lnstit., Davis, CA.
Taitel, Y. and Dukler, A. E., 1976, A model for predicting
flow regime transitions in horizontal and near horizontal
gas-liquid flow. A. I. Ch. E. J. 22, 47-55.
Tam, V. H. Y. and Cowley, L. T., 1988, Consequences of
pressurised LPG releases: The Isle of Grain full scale
experiments. Proceedings of the GASTECH88 Conference,
Kuala Lumpur.
Trapp, J. A. and Ransom, V. H., 1982, A choked-flow calcu-
lation criterion for nonhomogeneous, nonequilibrium,
two-phase flows. Int. J. Multiphase Flow 8, 669-681.
Zielke, W., 1968, Frequency-dependent friction in transient
pipe flow. ASME J. Fluid Engng 90, 109-115.

You might also like