RESEARCH
CENTRE
EUROPEAN COMMISSION
Institute of Safety Technology
Neither the Eur opean Commission nor any per son acting on behalf of
the Commission is r esponsible for the use which might be made of the
following information.
Catalogue: CL-NA-15951-EN-C
Printed in Italy
Abstract
This report describes a steady-state Homogeneous Equilibrium Model (HEM) that can be
applied to the design of emergency pressure relief systems involving compressible flashing
flows. The relief system may comprise an inlet nozzle (between vessel and vent-pipe inlet), an
inlet pipe with various fittings (bends.etc), a safety valve and a tailpipe with fittings including
area changes. Multiple choking may occur in such systems and the model includes a procedure
for dealing with this situation.
The equations of the model have been assembled into a computer program called VLINE and
this program is applied to a practical relief system to illustrate its usefulness in design.
Contents
Nomenclature 1
1 Introduction 7
8 Discharge coefficients 17
10.1 Derivation of differential equation for local void fraction in vertical vessels of
constant cross section area 21
10.2.1 Definitions 22
10.2.2 The DIERS correlations for the drift flux and the superficial velocity of
the gaseous phase 24
11 Computer program 29
33
12 Conclusions
34
References
37
Appendix A
39
Appendix
Nomenclature
Physical property constants defined by Equation (6)
Local physical property constant defined by Equation
(27)
Local constant defined by Equation (32)
b Physical property constant defined by Equation (5) -
Cd Discharge coefficient -
D Diameter m
f Friction factor -
Fa Function of ^ -
Greek
Letters
Void fraction -
Viscosity Pas
Density kg/m3
Surface tension N/m
Parameter of Leung and Grolmes (Equation (10) -
In the present work, a steady-state Homogeneous Equilibrium Model (HEM) has been
developed for sizing relief lines under compressible flashing flow conditions. The model for the
relief line represents an extension of the work of Morris [1] which did not include a safety
valve model. Additionally, in order to obtain a complete model that is also able to describe
approximately the conditions in the vessel, parts of the DIERS technology [2] have been
reviewed and some new results are presented. The equations for two specific vent-line
configurations and for a vertical vessel have been assembled into a computer program. One of
these relief systems is shown in Figure 1.
Since many industrial relief systems are complex networks of pipes comprising many fittings
such as bends, valves and area changes, some concessions are made to simplicity so that the
model for the vent line can be easily used by industry for the design of arbitrary relief system
configurations.
An additional argument for simplicity is the fact that it should be possible to use the relief line
model as a boundary condition for the computer code RELIEF [3] which is actually under
development at the JRC. This code models the transient hydrodynamic, thermodynamic and
kinetic phenomena of multi-component two-phase mixtures within a vertical vessel. In order to
make the code user-friendly,time-consumingcalculations have to be avoided where possible,
i.e. a simple vent line model must be used. This choice is justified since the behavior within the
vessel determines the entrance conditions to the vent line, and the errors associated with
calculating these conditions usually have a much greater effect on the calculated ventflowthan
the likely spread of values that would result from using more sophisticated models for the relief
line.
Atmosphere
Tail Pipe
(Inner Diamter D2 > Dl)
Inlet Tube
(Inner Diamter Dl)
Inlet Nozzle
Vessel
Fig. 1: Typical relief system configuration
The great advantage of this approach is its ability of modelling nonequilibrium effects in detail
and its general character which makes it applicable to a large variety of problems. Examples of
internationally well-recognized computer codes based on this method are the US codes
RELAP5 (INEL) and TRAC (LANL), and the French code CATHARE (CENG) which have
been developed for the analysis of Light Water Reactors (LWR) under off-normal and accident
conditions. The comparisons of predicted and measured data show that the two-fluid model
can be an efficient tool for simulating two-phase phenomena. However, two major
disadvantages have to be taken into account :
A complex differential equation system must be handled, i.e. long calculation times and
numerical problems may occur
Detailed information is needed both for the interactions between the phases and for the
initial and boundary conditions (e.g. information on flow regime, interphase structure,
initial nucleation site density)
Therefore, to attain the scopes defined in paragraph 1, the following simplifying assumptions
are made:
1. The two phases are in mechanical, thermodynamic and kinetic equilibrium, i. e. PL = PG >
T
L = TG a n d U L = UG
2. Steady state conditions in the vent-line are reached rapidly
With the first assumption the two-phase mixture can be treated as a pseudofluid with suitable
average properties that obeys the equations of single-phase flow [4] This approach is known as
the Homogeneous Equilibrium Model (HEM). It gives good results when there is sufficient
time for thermodynamic equilibrium to be achieved and when the flow pattern is conducive to
interphase forces that are adequate to supress relative motion [5] This is normally the case in
long pipes at high velocities and low qualities when bubbles are entrained in a turbulent flow.
The assumption of an isenthalpic expansion law is generally justified for frictional flow in a
horizontal pipe during relief conditions where the residence time of the fluid is too short for
appreciable heat transfer and where the changes of kinetic and potential energies are small
enough to be neglected. Again this is only an approximation in the case of tube-
reductions/enlargements where the change of kinetic energy plays an important role and where
the expansion law is better described by an isentropic path. Furthermore, in complex networks
with very long relief lines (e.g. relief systems on offshore oil platforms [6], [7]) heat transfer
might become important and a deviation from an isenthalpic path may be appreciable.
By assuming an isenthalpic expansion law one avoids the need of solving the momentum and
energy balances simultaneously. The steady-state momentum equation for one-dimensional
homogeneous two-phase flow in a duct with non-constant cross-sectional area and negligible
gravitational pressure drops may be written 4:
. . 4ftDdz 2x
dp = -md(mv h ) ^ - m vh (2)
10
where ftp is the two-phase friction factor, rh is the mass flux (M/S] and v^ is the
homogeneous specific volume defined by
v
h=xvG+(1"x)vL (3)
The mass flow quality is equal to the thermodynamic quality since equal phase velocities are
assumed. As the thermodynamic path is known and the phases are supposed to be in
equilibrium, the quality and the homogeneous specific volume v^ are only functions of
pressure. Equation (2) can then be integrated numerically by executing flash calculations for
each pressure drop increment At each step the thermodynamic properties must be taken from
Steam Tables or from equations of state. This is still atime-consumingprocess that can be
avoided by making use of the DIERS result [8]
v2
vh , 1-L ,
- = l+a + b^ 1 (4)
v
ho J
where = p/p 0 is the pressure ratio, ^ 0 and p 0 are known upstream reference values and a
and b are constants. Equation (4) is an approximation of the homogeneous specific volume
along an isenthalpic path in the two-phase region between upstream and downstream
conditions and allows the analytical integration of Equation (2).
The constants a and b can be obtained from Equation (4) by two isenthalpic flash calculations
at some known downstream pressure P2 (e.g. pressure in a catch-tank) and some intermediate
pressure p ! = ( p 0 + p 2 )/2
a= - ^ - 1 ^ - 1 ) (6)
1
where
A=-^- (7)
v
ho
and
Leung [9] and Leung and Grolmes [10] have reported that the following equation,
11
^ = 1+1 (9)
v
ho
where
x = (Avho-vL)/(vG-vL) (H)
Some means is required by which VQ and VL can be related to upstream values VQ 0 and VL 0 .
It is proposed here that the vapour expands polytropically (pv = constant) between the
upstream and downstream pressures (p 0 and P2), in which case the polytropic index n can be
found from
A further reasonable assumption is that the liquid phase is incompressible (i.e., VL = VLO) in
which case Equation (11) becomes
Using Equations (3),(7) and (14), the homogeneous void-fraction may be found from
a h = l-(l-x)vLo/(v h o A) (15)
solving the momentum equation for a nonconstant crosssection area. Especially for
enlargements it is inadvisible to apply the equivalent length method since the enlargements are
possible choke locations.
The relief line is therefore treated as a series of straight, horizontal pipe sections of different
diameters connected to elements with diameter change.
A=1 + + b\
Atmosphere
pipei I
I
7
yV^l+a-^t+biXi
A = l + k + b'kl (16)
Equation (16) then describes the variation of twophase specific volume in terms of pressure as
the fluid flashes between the pressure vessel and the catchtank or the atmosphere.
In order to facilitate the numerical treatment by subroutines it is desirable to have a local form
of equation (16) e.g.
A i = l + a ^ i + b i \f (17)
ki=(lTli)/Tli (19)
and
T (20)
li=P/Pio
The constants a and bj will, in general, differ from the known constants a and b. However,
noting that >
A = A / A 10 (21)
where
13
A
io = vhio / vho = 1 + 3 +
(22)
=(1-)/ (23)
=/ (24)
and making use of the relationships
-=/ (25)
+1=(+2)/-2 (26)
it can be shown that the constants aj and b in Equation (17) can be related directly to the
known constants a and b through the equations,
a = a / ( ^ 0 A 0 ) + 2 b ( l ^ 0 ) / ( T A i o ) (27)
and
Thus, a and b[ can be found by specifying only the inlet pressure pi 0 to the ith pipesection.
Integrating this equation along the pipe (Length Lj) assuming a mean value of friction factor
ftp results in the following working form of the momentum equation,
where
Ai=(ai2bi)/(aibil)2 (32)
14
B i =l/(l + b i a i ) (33)
di=)/(a24bi) (35)
Equation (30) must be solved iteratively for the pressure ratio Tjj that satisfies the equation for
given values of pj 0 , m, Lj and Dj. The pipe exit pressure j e is then given by
Pie=TliPio (36)
For subsonic flow, pj e can be taken as the inlet pressure to the next pipe section and the
process repeated.
Retp=^ (37)
where
is the twophase viscosity. The quantity a^ is the homogeneous flow void fraction.
In the present work, a mean twophase friction factor is required for use in Equation (30).
Morris [1] proposed that this be based on a void fractionaveraged Reynolds number Retp
defined by
, a,*
Re =
KTJ R e , p < t a l '
a
ho
<39>
where a ^ and oche are the pipe inlet and outlet void fractions respectively. Equations (37)
(39) then give
15
where
s e = a h e - 0.2(1.5 + 0 / (42)
and
Morris [1] recommended to determine the two-phase friction factor ftp from the following
form of the Chen equation [12] for commercial steel pipe:
where
9rh*
=0 (46)
<fr|i
=
Applying this criterion to Equation (30) yields the following equation which must be solved
iteratively for T|jc :
(2b^ic+ai-di)(ai+di)
Y = A i l n f o c V i c ) + B i ( l - r i i c ) + Ciln< (48)
(2biXc+ai+d)(ai-d)
16
A short way of deriving Equations (46) and (48) is given in the Appendix A.
The choking pressure pic at the exit of the ith pipe will be given by
Pic=TlicPio ( 49 )
The choking mass flux mc* can be readily found from Equation (47) by replacing Y with the
l.h.s. of Equation (30)
Note that this equation corresponds to the result from Leung and Grolmes [10] when setting aj
= and bj = 0.
Mi=CdMd (51)
The ideal mass flowrate is calculated from the momentum balance. For frictionless flow
Equation (2) may be rewritten in the following form
AidTi^ldtm2^) (52)
Integrating this equation from the inlet to the throat of the nozzle yields the following working
form of the momentum equation for an ideal nozzle
M
i,id
m* > h io /Pio) (55)
Si,
17
8 Discharge coefficients
Equation (51) defines the discharge coefficient Qjj. It can be obtained by measuring the real
mass flowrate through a restriction for an imposed pressure drop and deviding it by the
corresponding ideal flowrate. In this case, Qy includes all the phenomena that have been
neglected in the ideal model such as friction, differing phase velocities and thermal
nonequilibrium. Therefore, the two-phase discharge coefficient depends not only on the
geometry of the restriction and on theflowrate,but also on upstream conditions like pressure,
quality and flow regime. Obviously a large amount of data is needed to obtain a useful
correlation for the discharge coefficient In most cases, these data are not available. It is
therefore proposed here to adopt a similar strategy like for the friction factor in the tube, i.e. to
deduce an expression from incompressible single-flow methods which includes the two-phase
Reynolds number defined in Equations (37) and (38).
The "two-K method" proposed by Hooper [13] predicts the pressure drop across pipe fittings
for single-phase flow. The advantage of this method is its applicability both to laminar and to
turbulent flow. The total pressure drop across a pipe fitting is defined as
Aptot=K-|pu2 (57)
and
K = K
turb + Klam (58)
where K tur b is only a function of the geometry and K ^ is inversely proportional to the
Reynolds number.
As follows, relations between the discharge coefficient and the K factor will be derived which
are then used to define suitable discharge coefficients for valves and inlet nozzles.
Aptot=K.|pu2 (59)
For incompressible single-phase flow, the total pressure drop from the inlet to the throat of a
valve is defined by
A
Ptot=Po-Pt + 2 p ( u o - u ? ) (60)
By introducing Ap tot from Equation (59) in this equation and by making use of the mass
continuity
18
ut = u 0 / (61)
m0=pu0 (62)
the following expression for the real mass flux at the valve inlet is obtained
2_ 2p(p 0 p t )
(63)
"(l- 2 )/ 2 + K
2 2p(p0pt)
m
o,id=-
^ (64)
According to the definition in Equation (51), the discharge coefficient is found by dividing the
real mass flux from Equation (63) by the ideal mass flux from Equation (64):
qi=i/[i+K2/(i2)] (65)
Now, the factor must be determined. By considering experimental data for a sudden pipe
contraction, in the case of laminarflow,it can be shown that, approximately,
32 1'
l a m (66)
~ R e 2
For a valve, the discharge coefficient is normally given for fullydeveloped turbulent flow. In
this case K j ^ can be neglected and Equation (65) can be used to calculate Kturjj from the
corresponding Qjtuflv
V
K
turb -
1
1 i- 2\ (67)
^Cd.turb
1
C
d,tp- (68)
l/cl,urb + 32/Re0ftp
where the twophase Reynolds number Re o t p refers to the real mass flux at the valve inlet
When the real mass flux is not known, Equations (37),(51) and (68) lead to a quadratic
equation for M that can be solved analytically.
19
In the case of the inlet from a large vessel to the ventline, the velocity in the vessel can be
neglected and the K factor refers to the velocity ut in the throat of the nozzle:
APtot=K.|pu2 (69)
A
1 2
Ptot=PoPt2put (7
)
The real and the ideal massfluxesin the throat of the nozzle are then defined by
m2 = 2 p ( p P t ) (71)
1
K+l
m 2 id=2p(PoPt) (72)
which leads to the following definition of the discharge coefficient for the inlet nozzle
OJ =! (73)
K+l
Introducing the factor defined by Hooper [13] into Equation (73) yields the twophase
discharge coefficient for an inlet nozzle
C
-"=i.5+m/Rc,,v (74)
where the twophase Reynolds number R e ^ refers to the real mass flux in the throat of the
nozzle. Like in paragraph 8.1, for an unknown mass flux, Equations (37),(51) and (74) lead to
a quadratic equation for Mj that can be solved analytically.
For that reason, in case of choked flow, it is sufficient to consider only the converging part of
the pipe fitting. In order to describe the choked two-phase flow across an ideal converging
nozzle, Equation (46) is applied to Equation (53) which gives the following expression for the
critical pressure ratio:
0=A?c-2 (75)
+2(a i +2b i X ic Xl + X ic ) 2 {A ic /(l + X i c ) - 2 b ^ i c - l - ( a i - 2 b i ) l n ( l + Xic)}
Equation (75) can be easily derived by means of the same method that has been used for the
tube and which is given in the Appendix A.
Again, Equation (75) has to be solved iteratively for T|jc. The choking pressure in the throat of
the fitting is then given by Equation (49).
Making use of Equation (53) and replacing Ajc - f in Equation (75) yields the choking mass
flux for the nozzle
m
ic* = T lic/-\/ a i + 2 b Aic (?6)
which is identical with Equation (50) for the tube.
One way of obtaining an accurate value for the entrance is to use the vent line equations as a
boundary condition for a detailed vessel model such as the RELIEF code. In the present work,
the DIERS technology [2, 8] has been used which allows the estimation of the entrance quality
during top-venting of a vertical vessel by hand-calculation methods. The basic assumptions and
equations are outlined in this section.
21
10.1 D erivation of differential equation for local void fraction n vertical vessels of
constant cross section area
As follows, the venting of a vessel of constant cross section area is considered. Initially, the
vessel is partially full of saturated liquid at pressure p. At some instant of time during the
swelling up process, an energy balance can be established for a fixed elemental volume of
height dH .The following assumptions are made:
2. The net liquid flow to/from the element can be neglected, i.e. JL = 0
3. The pressure gradient within the vessel due to friction and hydrostatic head is negligible
where is the local void fraction at height which is averaged over the cross section area.
Combining and expanding of Equations (77) and (78) and making the approximations
and
0 / ~ ) = (1-)* (82)
where the terminal rise velocity of a single bubble UM is introduced in order to non-
dimensionalise the superficial vapour velocity JQ and where
2
_ ~vLGhLHo (g 3 )
v
LvGhLGUo
and
*
= (84)
Hn
The next important step is now to find a simple relation between the superficial vapour velocity
JG and the local void fraction so that Equation (82) can be integrated analitically over the
non-dimensionalised mixture height H*. This relation is provided by the drift flux theory which
is presented in the following section.
For the flow in large vessels, it is a reasonable approximation to neglect acceleration and wall
friction forces. Wallis [4] has shown that in this case, in which gravity or some other body
force is balanced by the pressure gradient and the interfacial forces, the relative motion of the
phases depends only on the void fraction and the system properties. The relative motion is
usually described by the drift flux which is defined below.
10.2.1 Definitions
The volumetricfluxesof the phases which are often referred to as the superficial velocities are
given by
J L = Q L / S = (1-CC)U L (85)
JG=QG/S = au G (86)
where Q L and QQ represent the volumetric flowrates and UL and UQ stand for the absolute
velocities of the phases. The total volumetricfluxis defined as
23
J = QLQG=JL+JG (87)
The d rift velocities are defined as the difference between the component velocities and the
average as follows:
uLj = u L j (88)
uGj=uGj (89)
Finally, the d rift flux represents the volumetric flux of a component relative to a surface
moving at the average velocity, i.e.,
JGL=a(u G j) (90)
JLG=0cO(u L j) (91)
The definitions indicated above are valid for uniform void fraction and velocity profiles over
the cross section. For nonuniform profiles, averages over the cross section are defined by
[XdS
(X> = ^ _ (92)
where the brackets < > denote averages over the cross section. Usually, only these averages
are obtained from experimental investigations. For example, the averaged superficial gas
velocity <JQ> in a bubble column is obtained by dividing the measured volumetric gas flowrate
by the cross section area. The averaged void fraction <co can be obtained by techniques or
by a "mouse trap" where a certain part of a test section is suddenly isolated by two
simultaneously closing valves.When introducing these averages in the above definitions, it has
to be kept in mind that the product of averages generally differs from the average of a product,
i.e.
<X).<Y)*(X.Y) (93)
Therefore, Zuber and Findlay [15] introduced the distribution parameter C 0 and the weighted
mean drift velocity uGj
C =-&L (94)
and accounts for nonuniform flow and void fraction distributions over the cross section. By
assuming several different flow and void fraction profiles, it could be shown that for fully
established axisymmetric twophaseflowwith a wa j < ct c e n t e r C 0 varies between 1 and 1.5.
24
and accounts for variations of the local relative velocities of the phases over the cross section.
10.2.2 The DIERS correlations for the drift flux and the superficial velocity of the gaseous phase
When correlating the drift flux in Equation (90), DIERS distinguishes between three vessel
flow regimes: Homogeneous, bubbly and churn-turbulent flow. In the first case, the vapour
bubble rise velocity relative to the liquid is assumed to be zero, i.e.
JGL.h = 0 (96)
Obviously, this assumption leads to an entrance quality which will be the same as the average
vapour mass fraction in the vessel. This model may approximate vessel conditions when the
vessel contents are extremely viscous or foamy or when the venting time is too short for
appreciable bubble rise [2]. For high vessel fillings, for example, the mixture level reaches the
top of the vessel rapidly so that the bubbles do not have the time neither to move from the
location where they were created nor to coalesce.
The bubbly flow model assumes a continuous liquid phase with discrete bubbles, whereas the
churn-turbulent vessel model considers a continuous liquid phase with coalesced vapor regions
of increased size relative to the bubbly flow model. Table 1 lists the drift flux correlations used
by DIERS to describe bubbly and churn-turbulent flow. The indicated quantities are averaged
over the cross section.
The DIERS drift flux correlation for bubbly flow is similar to the local correlation form (i.e.
non-averaged over the cross section) proposed by Wallis [4] or Zuber and Hench [16]:
jGL=U00-a(l-a)k * (97)
Equation (97) is modified in the DIERS report [8] by introducing the denominator (l-cc^),
probably in order to facilitate subsequent integral analisys (the motivation for this modification
is not documented). In fact, the denominator has little effect for a<0.4.
The expression used for the terminal rise velocity during bubbly flow has been found by
Peebles and Garber [17] with the only difference that DIERS accounts for the influence of the
density of the lighter phase which is neglected in [17]. It is noteworthy that both in the bubbly
and in the churn-turbulent flow regime the terminal bubble rise velocities are independent of
the bubble diameter and viscosity.
The drift flux correlation for churn-turbulent flow is based on the work of Zuber and Hench
[16] who found that the drift velocity of the gaseous phase in a churn-turbulent flow regime is
independent of the void fraction, i.e..
25
u G j=U 00 (98)
Using this result together with the Equations (89) and (90) yields the drift flux correlation for
churn-turbulent flow in Table 1. Note that, since the terminal rise velocity is a constant, this
correlation is valid both for averaged and non-averaged quantities.
The expression used for the terminal rise velocity differs from that for bubbly flow only in the
constant equal to 1.53 which wasfirstproposed by Harmathy [18].
The local drift flux is related to the superficial velocity of the gaseous phase by Equations (86)
and (90), i.e.
JGL=JG-J' (99)
The averaged form of this equation is
(JGLHJG)-H (10)
With the definition of the distribution parameter C 0 from Equation (94) the averaged drift flux
is given by
The quantities in the following equations are all averaged and the brackets are omitted for
convience.
jo = JGL+C<XJL (102)
1 CCL0
Neglecting the superficial velocity of the liquid phase JL and introducing the DIERS drift flux
correlations for bubbly and churn-turbulent flow respectively leads to the expressions for the
superficial velocities of the gaseous phase in Table 1.
26
Bubbly Churn-turbulent
(1-) 2
JUL - 3 JGL = u ~ a
1 = 1.53[Ggv L (lv L /v G )] V4
U 00 =1.18-[agv L (l-v L /v G )] 1/4
_y (1-) 2 JG=Uo
(lC0a)
G 3
(l-C0a)(l-a )
Tab. 1: DIERS correlations for the drift flux, the terminal bubblerisevelocities and the superficial velocities
of the gaseous phase
With the superficial velocities JQ for the gaseous phase as defined in Table 1 it is now possible
to integrate Equation (82) in order to obtain an expression for the void fraction at the top of
the vessel. This procedure is described exemplarily for churnturbulent flow.
Replacing JQ in Equation (82) with the equation from Table 1 for churnturbulent flow leads to
da
= 7 (103)
(la)(lC 0 ar
Equation (103) may be integrated over the mixture height
an da
J (104)
(la)(lC 0 a)'
where H* m a x stands for the height at the mixture surface and a.max denotes the void fraction
at this location.
In addition to Equation (104), the definition for the average void fraction of the whole two
phase mixture offers a further relationship between the nondimensionalised mixture level H*
and the void fraction which is averaged over the cross section at a certain height of the
vessel:
"max
I f *
= 51 j adH (105)
H ITIT
max o
By making use of this definition, the unknown variable can be eliminated from Equation
(104).
= V G /V = (VV L )/V
= ( H H 0 ) / H = 11/H
Equation (106) is valid for vessels with constant cross sections and it is assumed that the
volume of the liquid phase remains constant during the swelling up process, i.e. the volume of
the evaporated liquid is neglected.
Equation (105) may be rewritten by replacing the differential dH* by means of Equation (103):
"max ~A~
= (107)
^- / *
(la)(lC 0ar
Integrating therighthandsides of Equations (104) and (107) and dividing yields
In [ l C 0 a m a x | | (C 0 l)a,nax
1 amax J \1 Coamax/ _ flos')
a = a a
t 1 ~ Cp max | ) V mc
max
a a
[ 1~ max J v.! ^o max/
This equation is quite cumbersome since the unknown quantity o ^ must be found by
iteration. Therefore DIERS approximated the void fraction at the surface of the twophase
mixture in the churnturbulent case with the following correlation
(109)
"S
The error that is introduced by this approximation can be evaluated easily on a computer since
the void fractions vary only between 0 and 1 and the distribution parameter lies between 1 and
1.5. The maximum error is found to be 3.1% for C0 = 1.5 and = 0.401.
Morris [19] repeated the same procedure for bubbly flow with the correlation from Table 1 for
the superficial velocity of the gaseous phase and obtained
=<xmaxX + Y (110)
X+Y
where
and
28
(Co-1) Co 1- max
Y= In i Z. ^max "*" -max In
+ Co + CoJ (l-ttmax) W 1 C0amax_
(112)
1 1 + Cf rV3-amax
2 arctan
V3 i + Cn+C
^ '"O 2 + amax /
where
5.75
m = ,1.85 (1-5) (114)
Note that this result differs from the DIERS result [2] which is given by
a
max (115)
Since the origin of Equation (115) is not documented, Equations (113) and (114) which have
been deduced from first principles are used in the present work.
In a last step, theflowingquality at the entrance of the vent line can be related to the maximum
void fraction at the top of the vessel by means of mass balances for the two phases and of the
drift flux correlation at the mixture surface. The resulting equation is the so called "coupling
equation".
JL.ma
max S = ( l X m a x ) M
V v vent (116)
L
JG.max
^v x
max^vent (117)
VG
where JL max anc* JG max a r e t n e superficial velocities at the mixture surface, Xmax is the
entrance quality of the vent line and M vent is the total mass flowrate in the vent line. The
superficial velocities are related via the drift flux. Applying Equation (101) to the mixture
surface and making use of Equation (87) yields:
where JGL,max denotes the drift flux at the mixture surface. Replacing JQLmax by the
correlations from Table 1 for bubbly and churnturbulent flow respectively and substituting the
superficial velocities by means of Equations (116) and (117) finally leads to the coupling
equations, i.e.
1 (lttmax) 2 . r VL
' 1_rv 3 ^v
T
x -c, vent * "max 1G_ 10
A u
max~ "max l L0amax / \ l 'J
1 V^G
for bubbly flow and
-+c^
x = a YG /J20
*max "max / \ \lvj
1 to a max v
v Gy
\2
_ m vent v G (121)
T
vent Uc
With the coupling equations, the quantities at the entrance of the vent line are defined and the
equation system is closed. For given stagnation conditions in the vessel (initial pressure and
initial filling) and a given backpressure (e.g. atmospheric pressure), the entrance quality, the
massflow and the pressure profile along the relief line can be calculated by iteration.
11 Computer program
All of the foregoing methods have been assembled into a computer code which can be used to
design various relief system configurations. The list of subroutines and the flow chart is given
in Appendix B. The flag MPMC is used to distinguish between the configurations of the
MPMC (MPMC = 1) and the COLUMBUS facilities (MPMC = 0) which are actually used for
blowdown experiments with water steam/mixtures. The C OLUMBUS configuration
corresponds to the typical relief system of Figure 1. The MPMC configuration is simpler, i.e. it
consists only of a vessel, an inlet nozzle and a straight tailpipe.
The example of Figure 1 (MPMC = 0) will be considered here to illustrate briefly the solution
procedure. The complete input data set is given in Figure 3. The program solves the equations
successively by forward calculations. Note that the procedure is based on the assumption that
choking occurs in the throat of the valve. This is realistic since industrial relief systems are
designed with defined choking locations.
Initially, the vessel is 8 5 % filled with saturated water. At 5 bar (absolute), the safety valve
opens and the vessel content is released to the atmosphere. C humturbulent vessel flow is
30
assumed. Contrary to the homogeneous case, the quality Xmax at the top of the vessel depends
on the mass flow which is not known a priori. Therefore, in order to obtain a first guess for
x
max t n e coupling equations are simplified assuming that
V
L/ V G (122)
max (123)
v
max (1OCmaxCo) *vent
where
(l<*max) (124)
Fa = a
1 max
Fa = l (125)
\2
a
max*a U D,
Miiax
(126)
m v D
max'-o/ vent G \ tJ
By assuming a critical pressure ratio T]c for choked flow in the throat of the valve (e.g. x\c =
0.6 which is used here and which is a good estimation for many design cases) and by neglecting
the pressure drop in the inlet nozzle and the inlet tube, the first guess for the mass flow
m v e n l can be estimated by means of Leungs' result [10]
m vent
= Jk_ [P~ ' Hvalve (127)
V Vho
where depends on the initial conditions and the quality at the top of the vessel
f 2
_ xmaxvLGo , C
LQTQPO
v
LGo j (128)
v n
ho vh0 LGoJ
The area ratio of the valve valve appears in Equation (127) since in Leungs' equation the
mass flux refers to the cross section of the throat of the valve.
Combining Equations (126) through (128) leads to the following approximation for the quality
v
max
where
=(<3->)/ (130)
c ^VGo-VLo^ (131)
2= c pLo T o
LhGo_nLo
a
max*a U0 D ,
c3 = (132)
( l - a m a x C o ) vGo valve
Note that both in the churn-turbulent and in the bubbly case, F a and a m a x depend only on the
initial filling and the distribution parameter C 0 . Therefore, the Equations (129) to (132)
express the quality at the top of the vessel entirely in terms of the initial conditions and are
used by the program to calculate the first guess for Xmax-
The estimated quality is now used to calculate the specific homogeneous volume at the top of
the vessel which in turn is needed to specify the coefficients a and b from Equations (5) and
(6). The first value for the mass flux can then be estimated by assuming choked flow through
the valve and by neglecting the influences of the inlet nozzle and the inlet tube. Equation (75)
for the choked pressure ratio in the valve is solved iteratively and the result is used in Equation
(50) to calculate the choked mass flux.
During the following steps, the mass flux is used to recalculate Xmax> vjj 0 , a, b and to
determine the pressure ratios of the inlet nozzle and the inlet tube by successive solution of the
momentum balances (53) and (30). The inlet pressure of the valve is then known and the
choked pressure ratio of the valve can be calculated. This result is used in turn to recalculate a
new choked mass flowrate. These steps are repeated until the difference between the new mass
flux and the previously calculated one is smaller than 0.001%.
At this stage, the mass flow is already well defined since the flow is supposed to be critical and
is therefore independent of the conditions in the tail-pipe.
For the tail-pipe calculations, two cases have to be considered. In the first case, which is the
most common one, the exit pressure of the pipe will be atmospheric, i.e. no choking occurs.
Solving the momentum equation (30) for the tail-pipe with the given mass flux then yields the
pressure ratio of the tail-pipe which together with the known exit pressure defines the back
pressure of the valve. However, it has to be checked if the calculated pressure ratio is a
physical solution, i.e. if the corresponding choked mass flow is larger than the actual mass
flow. If this is the case, the solution is found and the output routine is called.
If no physical solution is obtained, the second case, i.e. choking at the exit of the tail-pipe,
must be considered. This problem is solved by varying the inlet pressure of the tail-pipe until
that unique value is found for which the corresponding choked flowrate equals the previously
calculated one. Therefore, an outer iteration routine proposes inlet pressures for the tail-pipe
and an inner routine calculates successively the corresponding choked pressure ratios and
flowrates. Note that choking at the exit of the tail-pipe means that the maximum flowrate
32
corresponding to the inlet pressure is obtained and that the exit pressure is higher than the back
pressure, but the flowrate of the system is still determined by choking in the throat of the valve.
Finally, the pressure distribution along the inlet tube and the tailpipe is calculated. This is done
by increasing the tube length gradually and solving the momentum equation (30) at each step.
The corresponding equilibrium temperature is provided by the properties routine, the quality
and void fraction profiles can be calculated by means of Equations (13) and (15).
Figure 3 shows the calculated pressure profile of the considered example for assumed chum
turbulent flow in the vessel. It can be used to verify if the recommended backpressure on the
safety valve and the admissible pressures in the tubes are not exceeded. No choking occurs at
the exit of the tailpipe. However, if homogeneous vesselflow is assumed, a similar curve is
obtained but the calculated exit pressure is slighdy higher than the atmospheric pressure, i.e.
1.287 bar. This is due to the assumption of homogeneous vesselflow which means no vapour
disengagement in the vessel, i.e. a higher mass flowrate has to be handled by the pipe. In the
considered case, this is only possible at relatively high pressures in the tailpipe which results in
choking at the tube exit.
Equivalent Length
j ^ Inlet jioizle Cte Vilve _ . .
5
<! ->i
l< ' > '
>C
4.5- ^ inlet Pipe1 Length ^ ,1
Viive pressure'drop
due to acceleration
4 -
; ; >
3.5" -> <; Choked-prmure in the threat- of the valve:
ee 3" 1 1 1
2.5"
a Valve pressure drop
1
>
1 ( due to expansion waves
! 2- ! ! !
atmospheric
1.5- 'Backpressure j^ >f, exit pressure
(1.013 tar)
'of the valve 1.29 bu "-1
1 - ^ f I ^
t 1
0.5- -, "Equivalent length Bend' ", *
Till pipe
1 1 I 1 1
I
o -0 1 2 3 4 5 6 7 I
Length vent line [m]
Fig. 3: Pressure profile in a typical water/steam relief system. Initial filling 85%, initial pressure in the
vessel 5 bar, vessel diameter 1.6 m, area ratio of inlet nozzle 0.0041, area ratio of valve 0.501, turbulent valve
discharge coefficent C j turb = 091. inlet tube diameter 0.102 m, inlet tube length 1.829 m, tailpipe diameter
0.154 m, tailpipe length 2.439m, tube roughness 0.046 mm, distribution parameter C0 = 1.2.
The foregoing example makes clear that the possibility of multiple choking must be taken into
account, otherwise the valvebackpressure and the occurence of choking in the valve cannot be
verified for all cases.
33
Even more important than the pressure profiles are the data shown in Table 2 in which the
results for different vessel models are compared. The corresponding input data set is the same
than for Figure 3. It can be clearly seen that with increasing vapour disengagement in the vessel
(no disengagement for homogeneous flow, best disengagement for churn-turbulent flow) the
massflowrateis decreasing whereas the volumeflowrateleaving the vessel is increasing.
In the case of a runaway chemical reaction or an external fire, the design engineer has to make
sure that the volume flowrate is sufficient to compensate the expected maximum volume
production rate in the vessel. At the sametime,high volume flowrates (which are equivalent to
high vapourflowrates)efficiently cool the vessel contents and slow down the chemical reaction
and the increase in pressure. Therefore, the most conservative design method is the
homogeneous vessel-flow model which predicts volume flowrates about a factor ten smaller
than the flowrates calculated by the other models. This means that, in order to get the same
volume flowrates than from the other models, much larger diameters are required. This may
lead to strongly oversized relief lines with unnecessarily high costs. Since the homogeneous
vessel-model also predicts the highest mass flowrates, the necessary size of downstream
equipment such as catch tanks or liquid knockout drums may be also significantly over-
estimated. However, often sufficient experimental data to assure the applicability of only one
of the models are not available so that it is advisible for these cases to base the design on the
homogeneous vessel-flow model.
Vessel-flow model x
max M [kg/s] Q[m 3 /s]
12 Conclusions
A steady-state Homogeneous Equilibrium Model for the design of long relief systems is
presented. Together with simple models for the vessel-flow based on the DIERS-Technology,
it allows the calculation of the maximum mass and volume flowrates and the pressure
distribution with the corresponding equilibrium quantities along the vent-line. The possibility of
multiple choking is taken into account. The calculations may be accomplished on a normal PC
within a few seconds enabling the designer to easily check different relief system
configurations.
In a further step, the results from the model will be compared to experimental data from the
MPMC and COLUMBUS facilities. In parallel, the vent-line model will be used for the
transient vessel-code RELIEF.
34
References
I] Morris S. D.: Flashing flow through relief lines, pipe breaks and cracks. J. Loss Prev.
Process Ind. 3, 1990.
2] Fisher H.G. et al.: Emergency Relief System Design Using DIERS Technology.: The
Design Institute for Emergency Relief Systems (DIERS) Project Manual. American
Institute of Chemical Engineers, 1992.
3] Duffield J.S., Friz G. and Nijsing R.: The venting of peroxide solutions. In 27th European
Two-Phase Flow Group Meeting, Stockholm (Sweden), June 1992.
6] Crawley F.K. and Scott D.S. IChemESymp. Series No. 85, p.291, 1984.
8] Fauske & Associates, Inc.: Emergency relief systems for runaway chemical reactions and
storage vessels: A Summary of Multiphase Flow Methods. DIERS Report FAI/83-27,
1983.
II] Beattie D.R.H. and Whalley P.B. Int. J. Multiphase Flow 8, p.83,1982.
12] Chen N.H.: An explicit equation for friction factor in pipe. Ind. & Engng. Chem.
(Fundamentals) 18, 296,1979.
13] Hooper W.B.: The two-K method predicts head losses in pipe fittings. Chemical
Engineering 21, 96-100,1981.
14] Muir J.F. and Eichhorn R. In: Proc. Heat Transfer and Fluid Mechanics Institute,
Pasadena (USA), June 1963.
15] Zuber . and Findlay J.A.: Average volumetric concentration in two-phase flow systems.
Trans. ASME, J. Heat Transfer 87,453-468,1965.
16] Zuber . and Hench J.: Steady state and transient void fraction bubbling systems and their
operating limits. General Electric Co. Rept. No. 62GL100, 1962.
35
[17] Peebles F.N. and Garber HJ.: Studies on the motion of gas bubbles in liquids. Chemical
Engineering Progress 49, 88-97,1953.
[18] Harmathy T.: Velocity of large drops and bubbles in media of infinite and of restricted
extent. AIChE Journal 6, p. 281, 1960.
[19] Morris S.D.: The bubbly flow model. EURO-DIERS, The Hague (Netherlands), October
1993.
37
Appendix A
Derivation of the equation for the choked pressure ratio in the tube
1 d kj = dA} 2ftpdz
2
(A.1)
m?* (l+X) Ai A D
we obtain an expression for the mass flux through integrating along the tube
.
dXi
:1 +
m 2 ( )
i* = 7 0 = \ (A.2)
(lnA +2ftpL i /D )
=(1T|)/TI (A.3)
and
dm* dm
-(1 + X)2^E =0 (A.5)
\ dl-,
Applying Equation (A.5) to Equation (A.2) yields the following equation type
FG F G = 0 (A.6)
where
1C
F.J A
0 ( + 0 ) ic (A.7)
A lnh
=A /Ti c )U+ nBi ii ( l ^ i cU) +rC il Jn!(2b
, ( c VA 2 i0XC i c+a
i Li -d
I ni )(a
(2bi?iic+a+di)(a_di)
M Li +d
I ^i)^
"IC
(.8)
F =
IC l (l + X ic ) 2 A c J ( + 0 )
38
G =(lnAic+2ftpL/Di)=^(a+2b^ic) (A.IO)
KXC CA1
Inserting Equations (A.7) (A.IO) in Equation (A.6) and simplifying leads to Equation (47) in
paragraph 6 for the critical pressure ratio expressed in terms of X\c and T|jc
where
Note that the integral in Equation (A.7) is already known from the derivation of the
momentum equation in paragraph 4. Therefore, this method of finding the choked pressure
ratio is much lesstimeconsumingthan deriving the explicit expression for the mass flux with
respect to Tij.
The same principle can be applied in order to derive Equation (75) in paragraph 9 for the
choked pressure ratio in a nozzle.
39
Appendix
Main Program:
VLINE.F
Subroutines:
Ab.f
Bisec.f
Cdinli
Etactf
Etacv.f
Etatf
Etav.f
Friction.f
Fund.f
Func2.f
Func3.f
Func4.f
Func5.f
Func.f
Func7.f
Input2.f
Mascv.f
Masstf
Outcolum.f
Outmpmc.f
Pipei.f
Props.f
Qtop.f
Xguess.f
Flow Diagram VLINE
VLINE
Input -Jnput2
ZU Props
Cale, for a given p sa t
Properties at p 0 :
corresp. T sa t and
c
Lo' a O' T 0' v Go' *| properties in the
water/steam equil.
VLo.hLo.hGo
1
O
Properties at
PM = ( p B +p o)/ 2: -Props
v
GM'vLM'hLM'
h
GM
I
Properties at PR :
v
GB'vLBnLB' -Props
h
GB
1
Cale. x 0 from filling:
0=1-
ovlo
v
Oo( - o) + ot o v Lo
1 a
I
First guess for quality
at the top of the
vessel. Set valve = 1
if MPMC =1 (i.e.
choking at the tube
exit) XfiliESS
Cale. ccmax.xmax'
and m start.
Distinguish cases
homog.,ch.turb. and
bubbly flow.
Cale. vno,x0,a,b
AB
Usex m a x and
previously cale,
properties to cale.
vho,Xo.a,b. Dis
tinguish methods
Leung (B=0) and
Morris (B*0)
Polytr. Exponent:
_ ln(P 0 /P B )
n =
^(VGB/VGO)
I
If MPMC = 0
cale, equivalent
length of bend and
total length of tail
Cale, crosssection
areas
PIPEI
If MPMC = 0, recal Cale, inlet quantities
culate first guess for and constants:
mass flow from = Pio/Po Mo<
assumed choked flow vhio*>vhio.Ai'Bi
in valve, neglecting IfA,-Bi-l*0,
the prcss.drop in the calculate also
nozzle of the vessel AA,BB,DD,CCi
and in the inlet tube
Mascv Props
Cale, first ideal mass Cale, for a given p sa t
flux from choked flow corresp. T sa t and
eq. Then cale, real properties in the
mass flux with dis water/steam equil.
charge coeff. depend
ing on Reto at the in
let. Take viscosities
from prop.routine
_0JQ&_
If vessel flow is not Cale, ctfnax and
homogeneous, max a s a function of
i recale, first quality at H the mass flux, vessel
the top of the vessel stag.condi tions and
and then a and b geometry.
A"
T
i
Find the exit pressurc(* PIPF.I
of lhe inlet nozzle
BISEC*- *FIINC3 * ETAV ar C DfNL <* ^Props
Set param, for etav cali. CDINL for Calc. void fraction at Cale, for a given p sa t
calc.discharge coeff. the exit of the inlet corresp. T sa t and
from given mass flux nozzle with given r\\. properties in the
and T|i. Then take viscosities water/steam equil
from prop.rout. and
cale. Rctp a ' this
point with given mass
flux. Finally cale, the
discharge coeff. as a
function of Rem
no yes
JL FTAT
Cale, equiv. length of Cale, difference
gate valve and total between the actual
length of inlet pipe. mass flux and the one
Then call ETAT to obtained from the
cale, difference Wjgcneral mom.eq. for a
between the actual given .
mass flux and the one
obtained from general
mom.eq. for given .
FT A CT
Cale, for a given |
RHS of mom.eq. for
.c
choking at the exit of
apipe.
IfMPMC = l **|- no
yes
1 MASST
Calculate new mass Cale, mass flux from
flux corresp. to choked flow eq. for
choked in tube given \
MASCV
Cale, new mass flux Cale, first ideal mass
with choked in the flux from choked flow
valve eq. Then cale, real
mass flux with dis PROPS
charge coeff. depend
ing on Retp at the in
let. Take viscosities
from prop.routine
-Yes
FIINC5
Set =
In an outer iteration Pio ^valvePin.valve
loop, guess the -MSEC' Call PIPEI -PTPFI
inlet/outlet pressure
ratio of the valve. FIJNC6
Cale, then the Find corresp. exit Set param, and call
corresp. choked exit pressure from choked -BISRO- FRICTION ^FRTCTION*- "PROPvS
pressure and mass
flux. Continue
flow in the tube
1
Set fto2 = ftDi and ETACT
iterations until call ET ACT
choked mass flow is
identical with actual Call MASST and
one. calculate the diff. -MASST
between choked mass
flow and actual one.
srv
MPMC = 1 ? -Yes
-No
OIJTMPM FIINC7
Calculate output data Output of input data.
for MPMC Set coordinates of
Configuration measurements along -9BISRC <- 5 Set param, for ETAT ETAT
the tube. Cale.
quantities at these
points -PROPS
OUTMCOUHvl
Calculate output data Output of quantities
for Columbus at the top of the
configuration vessel and at the exit PROPS
of the inlet nozzle.
I
Cale, and output of -PIPF.T
quantities along
the inlet tube ->BJSE4- HFIJNC7 4- -FTAT
-PRQPS
Output of quantities
at valve inlet,throat PROPS
and valve exit
1
Cale, quantities along -PTPET
the tail-pipe. Output. -tBJSEO- -EIJIC2' -ETAT
-PROPS