Two-phase flow venting from reactor

vessels

J. C. Leung and H. G. Fisher*

Fauske & Associates, Inc., 16WO70 West 83rd Street, Burr Ridge, IL 60521, USA
* Union Carbide Corporation, South Charleston, WV 25303, USA

This pdper provides an overview of the emergency relief requirements for reactor vessels under
runaway conditions. The importance of utilizing appropriate thermokinetic data in providing the
necessary energy release rate and pressure generation rate is emphasized with experimental styrene
data. The effect of two-phase flow venting on reducing the volumetric flow capacity of a pressure
relief device is clearly elucidated. In addition, the paper presents a concise and unified treatment of
two-phase flashing flow in the form of formulae and design charts covering single component fluids,
multicomponent mixtures, nozzles, long pipes, elevation changes, and back pressure effects.
Finally, the effect of allowing for overpressure in lowering the required relief area during two-phase
venting relative to the no overpressure case can be quickly assessed via the proposed normalized
relief sizing formula.

(Keywords: two-phase flow; vents: flashing flow)

Emergency relief systems (ERS) are installed to protect identified above in terms of the energy release rate and
process vessels from the catastrophic effects of an the pressure generation rate.
excessive overpressure and subsequent rupture. The Characterization of the flow regime behaviour in both
purpose of this paper is to focus on this important ERS the vessel and the relief system.
technology particularly in view of the recent advances Safe disposal of the relief discharge when the material
upon completion of the Design Institute for E1yrgency is considered too hazardous to be vented directly to the
Relief Systems (DIERS) research program ’ Relief atmosphere.
devices, properly sized only for the vapour load, may
‘Ihe discussion in this paper will be centred around the
often be ineffective because of the inadequate disen-
second and third considerations. However, it is equally
gagement of the vapour from the liquid and the resul-
important to recognize that the first item is by far the
ting two-phase flow. Numerous articles3-5, have
most difficult to define, and in fact the whole design
appeared in recent years dealing with emergency relief
rests on the adequacy of this ‘worst case’ to cover all
sizing for two-phase flow situations. Our approach is
credible upset conditions.
not to illustrate the use of these various design methods,
but rather to build on the existing technology. Speci-
fically, we will present a unified treatment of two-phase Thermokinetic data
flashing flow in pressure relief systems and downstream
Thermal data in terms of temperature and pressure are
piping with particular reference to multicomponent
crucial in the relief sizing calculation for the case of
mixtures, frictional losses, elevation changes, and back
runaway chemical reactions. As will be discussed later,
pressure effects.
these data provide the equivalent vapour/gas volumetric
To begin, the proper and safe design of an ERS for
generation rates. The approach based on detailed
coping with an uncontrolled runaway reaction requires
modelling of reaction kinetics and vapour-liquid-
the following considerations:
equilibrium (VLE) is often both time-consuming and
l Identification of the worst credible upset scenario to cost ineffective. However, it does offer the versatility of
establish the design basis. being able to extend the data to related conditions that
l Thermal characterization of the runaway reaction may be necessary to evaluate process variable changes.
In general a runaway reaction can best be experimen-
tally simulated in an apparatus that has a low thermal
inertia and thus approaches the behaviour in a process
This paper was presented at the ASME Winter Annual Meeting,
vessel. Such a bench-scale apparatus has been developed
27 November-2 December 1988, Chicago, Illinois, USA
as part of DIERS research program6,‘, and is available
Received I7 JQN&U~ 1989 commercially under the name Vent Sizing Package
0950-4230/89/020078-09S3.00
Zl 1989 Butterwonh & Co. (Publishers) Ltd
78 J. Loss Prev. Process lnd., 7989, Vol2, April
 Two-phase flow venting from reactor vessels: J. C. Leung and H. G. Fisher

This temperature is apparatus-dependent (since instru-

ment sensitivity varies) and furthermore can be equip-
ment specific due to intrinsic heat losses.
0 Tempering temperature is the temperature level corres-
ponding to a particular system pressure whereby the
reaction energy release is totally removed via latent
heat of cooling by the volatile component(s) present.
The reaction is said to be tempered or controlled
without any further rise in temperature. Such a system
has come to be known as a tempered system.

Theory of venting relief

100 I30 160 190 230 270 320 380 440
Temperature PC,scole : I/TK) Pressure relief of a runaway reactor requires a balance
between the volumetric vapour/gas generation rate 8
Figure 1 Styrene runaway data in adiabatic apparatus of
different @-factor and the volumetric discharge flow. This criterion must
be satisfied at the turnaround point in pressure. (Codes
may require this turnaround peak pressure not to exceed
(VSP: trademark of Fauske & Associates Inc. and Fike 1.1 times the design pressure of a process vessel.) In
Corporation). other words, if the vapour generation rate is greater
The importance of using thermal data obtained in a than the discharge rate, the reactor pressure will con-
low thermal inertia (sometimes called @-factor*) test cell tinue to rise. If the pressure is not allowed to increase
or correcting the data to a low d-factor corresponding beyond the intended relief set pressure, i.e., a zero
to that of the process vessel can be explained with the overpressure situation, the required relief area is given
help of Figure I. Styrene runaway polymerization data
from a VSP (1.05 d-factor) and from an ARC (1.5 by
d-factor as reported previously’; ARC is trademark of Ao=a=i)p, (1)
Columbia Scientific Industries) are shown. The Guo G
difference is obvious in terms of the peak adiabatic where vo = I/PO. Here uo is the specific volume, po is the
temperature and the peak self-heat rate. More impor- density corresponding to the inlet stagnation condition,
tantly, the self-heat rates at any temperature cannot be and G is commonly known as the mass flux (or mass flow
related simply by the ratio of the two @-factors, because rate per unit area) which may correspond to a choking or
the conversion at the particular temperature must also critical flow condition depending on the prevailing down-
be accounted for. Data obtained with a +-factor of 1.5 stream pressure and the relief configuration.
can be corrected to a 1.0 or 1.05 &factor with some To limit the scope, the remaining discussion will be
success 3. This can be an onerous task, however, if the limited to tempered (vapour pressure) systems where
data exhibit more than one exotherm, i.e., more than flashing of the liquid phase is expected during relief both
one peak in the self-heat rate plot. Extrapolating the within the process vessel and in the relief system. In this
data beyond the original temperature range is usually case, the net volumetric generation rate (taking into
considered to be unacceptable due to the possibility of account the net volume change upon vaporization which
additional, higher temperature reactions. on a per unit mass basis is simply ufg) is governed by the
In general, the following data must be obtained to energy release rate from the reaction,
characterize a given runaway reaction.
Adiabatic temperature rise is related to the heat of (2)
reaction or total heats of reaction for more than one
exothermic reaction. where mo is the initial batch size, qs is the product of the
Self-heat rate is a measure of the reaction rate and the liquid specific heat C,, and the adiabatic self-heat rate
energy release rate on a per unit mass basis and can be measured at the temperature corresponding to the set
used to yield an equivalent vaporization rate necessary condition, vr, is the difference in specific volume
to remove the reaction heat via latent heat of cooling. between the vapour and the liquid, and hf, is the latent
Pressure rise rate provides another measure of the heat of vaporization. Thus Equation (1) can be rewrit-
reaction rate, but here it relates to how rapidly pressure ten as
is building without relief.
Onset temperature provides a temperature level where
Ao= .mG(d WWs.
Gvo
the runaway reaction is expected to be self-sustaining.
This equation, which can be derived from an energy
balance’, says that the reaction heat is removed via the
* The so-called +-factor is defined by (m.C, + w?bCb)/m,C, where m.
and rr& are the masses of the sample and container respectively, and latent heat of the volatile component(s), the so-called
C, and Cb are the specific heats of the sample and container tempering behaviour.

J. Loss Prev. Process Ind., 1989, Vol 2, April 79

Two-phase flow venting from reactor vessels: J. C. Leung and H. G. Fisher

flow, which is consistent with the assumption of a

volatile vapour system discussed earlier. The HEM
model tends to give the lowest flow rate and a conserva-
tive relief size. In the past, such a calculation was rather
time consuming and required extensive thermodynamic
properties.

Generalized correlation for nozzle

Recently generalized HEM correlations in the form of
both formulae and charts have been presented for
nozzle flow lo as well as horizontal pipe flow”. The only
correlating parameter is given by
2
XOUfg GTP uig
w=---+- -
(4)
vo vo (3 hf

where all properties are evaluated at the known

upstream stagnation conditions (note that subscript ‘0’
lcl,,,,,,,,lb Inlet void fraction
denoting inlet condition has been dropped in urs, C,,, T,
P, and hr, for convenience). The ability to evaluate
Figure 2 Variation of two-phase density (p,), mass velocity (G), two-phase choked flow based simply on the inlet prop-
and volumetric flow per unit area (G/p,) with inlet void fraction
erties is a most useful finding. Here uo is the inlet
two-phase specific volume given by ur + xovrg evaluated
Effect of two-phase flow on relief capacity at the stagnation condition. Note that the first term can
The capacity of a relief device is normally rated in terms be approximated by (Ye, the inlet stagnation void frac-
of a specific vapour volumetric flow rate. If such a tion. For low quality flow typical of homogeneous-
device, designed for vapour, is relieving a saturated vessel venting, the second term dominates and the first
liquid or low-quality two-phase flow, its volumetric term OLO can be ignored in Equation (4). The generalized
capacity will be greatly reduced. According to Equation correlation for nozzle flow is presented in Figure 3
(3), the term Gvo, which can be thought of as a where the normalized critical mass flux G,,/fi (or G *
volumetric flow per unit area, plays an important role in a dimensionless quantity) and the critical pressure ratio
determining the relief size. It is therefore of interest to PC/P (or voc) are correlated solely in terms of w. This is
examine this term for a wide range of two-phase inlet a more compact form than a recently proposed
conditions. method “, in which both G* and not were correlated in
Using styrene properties at 500 kPa, Figure 2 shows terms of two parameters, the inlet quality xo and the
that both the mass flux G and the two-phase density PO so-called ‘flash parameter’ given by - P(dx/dP).
decrease as the inlet void fraction 0~0 increases. In
contrast, the volumetric flow term GVDor G/p, is found Multicomponent systems
to increase as (~0 increases. In making these evaluations, Although the correlation in Figure 3 was originally
the classical homogeneous equilibrium flashing flow
model (HEM) was used. We find in this example that
the volumetric flow per unit area is about 20 times larger
for all-vapour venting (a0 = 1) than for all-liqujd
venting ((~0 = 0). In other words, for the no overpressure
case, the required relief area is 20 times larger for
all-liquid venting than for all-vapour venting. Since
many reactors are typically operated nearly full of
liquid, the penalty for two-phase venting is substantial if
overpressure is not permitted under the conservative
assumption of homogeneous (uniform froth) venting.
We will return to the discussion on overpressure
behaviour as it relates to the relief requirement, but first
we will present the unified two-phase flow methodology
which plays such an important role in the relief sizing
evaluation.

Theory of two-phase 5ow

DIERS methodology’ recommends the use of the HEM
model in the relief discharge calculations. Again, the Figure 3 Generalized correlation for flashing choked flow
present discussion will be limited to flashing two-phase through a perfect nozzle

80 J. Loss Rev. Process Ind., 7989, Vof 2, April

 Two-phase flow venting from reactor vessels: J. C. Leung and H. G. Fisher

developed for single-component fluids, extension to

multicomponent systems is straightforward. The
additional information needed is the vapour composi-
tion. In this case, the representative latent heat of
vaporization may be approximated by
hfg = C Yihfgi (5)

where Yi is the vapour mass fraction of the ith com-

ponent. Likewise
Ufg= C YiUfgi (6)

Note that these values correspond to upstream stag-

Figure 4 Generalized correlation for flashing choked flow in a
nation conditions, which are generally known. The horizontal pipe
mixture liquid specific heat is simply the mass weighted
average property, i.e., :: I.0
r?
C, = CX,Cpi 0”
(7) ‘1 0.8
Y
To demonstrate the applicability of this correlation for E
a multicomponent system, a four-component mixture .B 0.6
z
with a wide range of volatility was selected as shown in &
Table 1. % o.sc Y
For simplicity, ideal solutions (obeying Dalton’s and
Raoult’s laws) were assumed throughout the calcula-
tions. ‘Ihe results for five mixtures with widely differing
compositions are shown in Table 2. For this particular I I lItIll 1 I I
illustration, the stagnation temperature is purposely 40 7 0 100 200 400
Equivolantlength/dtimeter
fixed so that the component saturation properties are
Figure 5 Comparison between DIERS vent flow data and correl-
unaltered. Also shown for comparison are the results ation: 0, ICRE-32 styrene (top); 0 ICRE-32 styrene (bottom); * ,
obtained from the DIERS computer code, V-32 ethylbenzene (bottom); q , IC’RE-2000 styrene Itop)
SAFIRE 13*‘4, which makes use of rigorous flash calcu-
lations. The proposed correlation yields results that are Figure 4 (Ref. 11). This discharge coefficient is simply
within 2% of the detailed calculations, thus lending the ratio of the flow in a pipe relative to that in a nozzle,
support to its application for multicomponent mixtures. i.e., G,/G,,.
Here the subscript c denotes the exit choking condi-
Effect of pipe length tion. If a constant two-phase friction factor of 0.005
The correlation for flow in a horizontal pipe has been (fully turbulent regime) is assumed, the results can be
presented in the form of a critical discharge coefficient presented in terms of the equivalent L/D of the vent line
(for exit choking) C, versus the flow resistance factor N as illustrated in Figure 5. Also shown are the experi-
(or 4fL/D) with w as the sole parameter as shown in mental DIERS data’531”, obtained with reacting styrem?
as well as non-reacting ethylbenzene (which has prop-
erties similar to styrene). The bottom-vented tests with
Table 1 Component physical properties at 120 OC
near zero inlet quality correspond to an w value of about
P hc C0 25. The top-vented tests exhibited an increased char-
Component M, IkPa) (kJkg_‘) hn3& (Jkg-‘Km’) acteristic disengagement that yields somewhat lo’wer w
water 18 205 2190.0 0.876 4240 values. Indeed it is significant to note that the data are
Ethylene glycol 62 5.7 972.8 9.379 2880 well bounded by the liquid inlet regime (w - 20) and the
Ethyl alcohol 46 427 754.5 0.157 1910 vapour inlet regime (k = 1.1 of Shapiro’s” chart). Also
Methyl alcohol 32 635 935.7 0.150 1620
note that the discharge coefficient is not very sensitive to

Table 2 Multicomponent flashing flow prediction TO = 120 ‘C,PO = 682 kg mm3

Composition (wt) Calc. G (kgm-*s-‘) GCDrrel

PO
EG EtOH MeOH (kPa) J,;;K ” carrel SAFIRE GS~FLRE

: 0.25
0.50 0.25
0.50 0.25
0.00 0.25
0.00 323.8
160.3 0.15
0.29 2660
3700 40.3
19.2 3050
1516 3100
1511 0.984
1.003

3 0.25 0.49 0.25 0.01 195.3 0.22 2970 30.1 1940 1960 0.990
4 0.25 0.25 0.45 0.05 273.1 0.15 2720 20.8 2690 2710 0.993
5 0.05 0.25 0.45 0.25 398.6 0.10 2200 13.6 3960 4020 0.985

EG, ethylene glycol

J. Loss Prev. Process In& 1989, Vol2, April 81

lkm-phasemw ventingfrom reactor vessels: J. C. Leung and H. G. Fisher

o in the range of 2@40 (typical values for most organics calculated ” using the formula
in the low quality region) as shown in Figures 4 and 5. Go (-2[wln 773+(w- l)(l- TJ~)])“’
Similar plots have also been presented by First and (9)
) PPO LOU/q3- 1) + 1
Huff l2 with xo instead of w as the correlating param-
eter. where Ga is the mass flux for the nozzle and 73 = P3/P,
i.e., the ratio of the back pressure Pa to the stagnation
E~ofekvationdmnge pressure P. It is convenient at this point to introduce a
To extend the above pipe flow correlation to include correction factor due to unchoked flow CF. Graphically,
elevation changes, an additional parameter in the form the results can be presented as CF versus
of (1 --~a)/l-noc) as shown in Figure 7. Once again the
additional correlating parameter is w. The critical values
Fr=gD-e p0gDc-e G,, and not are obtained from the corresponding nozzle
(8)
4fPvo 4fP correlations in Figure 3.
is required where 0 is the angle of inclination of the vent For the case of a vent line, the following approximate
line from the vertical. This is a modified Froude number formula has been found to yield reasonable results for
with the Fanning friction factor term 4 f incorporated in subcritical flow,
the denominator so as to allow use of the same value in
2(r1-9a- N Fr/ea)l I”
both N and Fr. F&m-e 6 presents such an extended (10)
NEa + 2(e3 - El)
correlation for situations commonly encountered in
reactor relief sizing (i.e., w ranging from 20 to 40). As and
shown, the higher the Froude number, the greater the
flow reduction or the smaller the critical discharge
p, = 773+ NC1 + W but limiting 71 not
l+N
coefficient at a given L/D. Note that this Froude to exceed 1.0. (11)
number does not explicitly contain the elevation change
ta = (VI + ?3)/2
term. But-for a given L/D and cos 0, a larger diameter
pipe would give rise to higher Froude number and larger 773 = P3lP

head term. According to its definition, a larger Froude 1 - Vi

number can be due to
0 A more vertically orientated pipe (smaller 0).
E,=W
( >
-
91
+1

where i in Equation (14) takes on values of 1, 3, and a,

0 A longer pipe (which for the same L/D would yield a and ci is the normalized specific volume Vi/U0 corres-
larger diameter). ponding to the present ratio vi. Equation (10) is derived
0 A lower relief pressure. from the momentum equation for a long duct while
0 A lower two-phase specific volume or a higher two- using a pressure-averaged value for the specific volume.
phase density The results based on Equation (10) can also be
All of these effects are adequately accounted for in the approximated by the curves in Figure 7 for the case of
present correlation. near horizontal vent line (Fr = 0). In this case, the
effective discharge coefficient Co for the vent line is
Efict of back presswe given by
The above discussion applies to critical (choked) flow CD = C,CF = G/G,, (15)
conditions at the exit of a constant diameter pipe. The
effect of back pressure on the relief capacity should be
considered particularly in the case of low set pressures
and long vent lines. In the case of a nozzle (such as a
safety relief valve), subcritical (unchoked) flow can be

Figure 6 Generalized correlation for flashing choked flow in a Figure 7 Generalized correlation for flashing flow in nozzles and
pipe with elevation change horizontal pipes with back pressure effect (Fr = 01

82 J. Loss Prev. Process Ind., 1989, Vol2, April

 H. G. Fisher

where C, = G,/Go, and CF = G/G, as defined before.

Table3 Comparison of styrene and phenolic system properties
To use the chart in Figure 7, st set pressure
r)itiacorresponding to the vent line has to be
System
evaluated. This can be related to the critical discharge
coefficient C, G,/G,, as shown ino f L e u n fioperty styrene Phenolic
and Grolmes”. For most situations where w
P, (kPat 450 207
Qarn 20 to 40, a good approximation for
7-8 (Kl 489 393
V tm31 13.2 4.54
mo (kg) 9500 4000
lJcO.9Cc = 0.9GcjGoc C,IJkg ‘K ‘) 2320 2930
dT/dt (“Cmin~‘) 28 10
~Tz (Kl 28 14
vfg (m3kgm’J 0.0772 0.866
h,,(kJk$‘) 292 2502
LID and Froude number, C, is
W(-1 25 25
found from Figure 6. With 73 and qc [Equation (16)] B (6) 0.4 0.4
known, Figure 7 then yields CF for the case Fr = 0 and A0 (m’) 0.613 0.168

113> Q (unchoked).
Co
>V3%QXIO) approximate G/G, as

In evaluating the mass flux G for a long vent line after G

-= ps+y APE +o.5$ 1
(21)
CD has been obtained, the critical mass flux G,, for a G, 5 6

nozzle case is first obtained from Figure 3. Then the The Arrhenius nature of a chemical reaction can be
mass flux G is simply given by C&G,,. Again, Equation approximated by the following formula*,
(10) offers an equally acceptable alternative. The
example in the appendix illustrates the use of these :=;[I +exp(+J] (22)
methods in various situations.
where A Ta is the temperature increase needed to double
the reaction rate and AT is the over temperature
Venting with ova-pressure
corresponding to the selected overpressure AP, these
Leung’ presented a closed form relief sizing formula for latter two quantities are =lated via the Clausius-
the case of homogeneous-vessel venting applicable to a Clapeyron equation. Up0 stitution, a generalized
wide range of overpressure. The equation as repeated relief sizing equation is ot
below reduces exactly to Equation (3) ‘when no over-
1 + exp[w”‘PAP/PJ
pressure is permitted, +,,= (2 + AP/P,)[l + (wAp/M”21’ (23)

(17) Here we have introduced another dimensionless param-

A = G~((vlmo)(hr~,uc.5 + (C, A7’)o.5l2
eter 6 that contains the physical property group
where P&J/C, T, and the kinetic parameter A Tz,

~=o.5cp[(~)s+
($J (18) (24)

Here subscripts s and p denote the conditions corres- Equation (23) suggests that as long as the physical
ponding to the set pressure and the peak turnaround parameters w and 6 are similar in any two systems,
pressure, respectively. the normalized vent area AlAa variation with the
The effect of overpressure on the required relief area normalized overpressure AP/ P, will be identical.
can best be illustrated by transforming Equation (17) The above-mentioned similarity can be demonstrated
into dimensionless form. By making use of Equation (3) using the styrene and phenolic reaction systems dis-
and the definition for w, the following equation is cussed in Leung’. Here both ue (or V/me) and AT2 for
obtained the phenolic case are judiciously chosen so as to yield
the same 8 value as the styrene case. Although the
GA (&Is )
(19) individual physical properties tabulated in Table 3 are
G,Ao = [lt (wAP/P~,“*12
quite different from each other, both systems yield an w
where G, would denote the mass flux at the set pressure. value of 2.5 and a /3 value of 0.4. Figure 8 shows the
In this equation the first term 010 in the definition of w prediction according to Equation (23) compared with
has been Adropped thus the results based on numerical integration of the time-
2 dependent mass and energy equations. These so-called

* A factor In 2 was obtained inside the exponential term AT/AT2 in

By noting that G increases nearly linearly with absolute Equation (22). This was subsequently dropped to yield a better
pressure for compressible two-phase flow, one can approximation.

J. Loss Prev. Process Ind., 1989, Vof 2, April 83

Two-phase flow venting from reactor vessels: J. C. Leung and H. G. Fisher

O.! j7-

0.'

0.:

s
0.:

ReComnended
0. L- 0.I-

( )-
0 0. I 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 04 0.5
AP/P, AP/P,
Figure 8 Generalized relief sizing equation in comparison with Figure9 Generalized relief sizing equation in comparison with
exact numerical solutions for w = 25: 0, styrene reaction; 0, DIERS styrene relief data: 0, ICRE-32 (top); 0, ICRE-32 (bottom);
phenolic reaction; p, Equation (23) o, ICRE-2000 (top)

‘exact’ calculations gave similar behaviour for A/& homogeneous-vessel venting is shown in Figure 9 to
uersus APIPS as expected. Note that at higher APIPs, provide an adequate, but conservative correlation of the
the analytical equation tends to yield a somewhat larger DIERS experimental data. The discrepancy is mostly
A/&, hence a more conservative result. Also shown in due to the observed vapour-liquid disengagement
this figure is the overpressure response at other values behaviour, which results in lower overpressures than the
of p. The curve for 0 = 0 corresponds to a constant idealized (homogeneous vessel venting) case. It is of
self-heat rate case and here the vent area decreases interest to note that the bottom-venting tests with no
monotonically as overpressure is increased. However, vapour-liquid disengagement gave results nearly in per-
with a @ value of 1.0, which would correspond to a fect agreement with Equation (23).
doubling in rate for every 11 “C temperature increase
for the styrene system (instead of 28 “C used in the
Conclusion
numerical example), Equation (23) yields a minimum in
relief area versus overpressure. This is due entirely to Simple models have been provided for assessing ERS
the approximation of taking the averaged q to arrive at requirements for runaway reactions assuming two-
the closed form solution. Note that the assumption of a phase homogeneous venting from a vessel. The advan-
temperature dependent q at the start would result in a tage of allowing for overpressure above the relief set
non-linear differential energy equation, which cannot be pressure can be quickly quantified using the proposed
solved exactly. To avoid the unrealistic solution beyond relief sizing formula. This paper also summarizes and
the minimum area overpressure, it is recommended that extends the generalized two-phase flow methodology to
the minimum area solution be extended to the higher include multicomponent systems, elevation changes and
overpressures as shown in Figure 8 (Ref. 18). back pressure effects.
The above illustration has shown that Equation (23)
provides a rapid means of evaluating the effect of References
overpressure once the key parameters w and /3 are
found. This equation also points out that a more Fisher, H. G., Chewi. Eng. Prog. 1985, 81 (8). 33
Fisher, H. G., in ‘The DIERS Users Group (A Forum for
meaningful definition of overpressure would be relative Development/Dissemination of Emergency Relief System Design
to the relief set pressure P, in the absolute scale. Technology)‘, Paper presented to the 1988 Spring AIChE Meeting,
Currently the practice is to define the per cent over- New Orleans, LA, USA, 1988
Huff, 1. E., Plant~Operations Progress 1982, l(4), 211
pressure with respect to the gauge set pressure. Fauske, H. K.. Chem. Eng. Prog.. 1985, 81 (8), 53
Finally, a comparison can be made between Equation Leung, J. C., AIChE Journal, 1986, 32 (lo), 1622
(23) and the available DIERS reacting system data’s*16. Fauske, H. K. and Leung J. C., Ci~em. Eng. Prog. 1985.81 (4). 39
Leung, J. C., Fauske, H. K. and Fisher, H. G., Thermochimica
Both large-scale (2000 1) and small-scale (32 1) styrene Acta 1986, 104. 13
runaway experiments were conducted with relief Whiting, L. F. and Tou, J. C., J.Thermot
Analysis, 1982, 24, 111

pressures around 550 kPa. The effective vent area incor- Fauske & Associates, Inc. in ‘Emergency Relief Systems for
Runaway Chemical Reactions and Storage Vessels: A Summary of
porates the flow reduction factor due to a long vent line Multiphase Flow Methods’, Technology Summary, DlERS/
according to Figure 5. Equation (23), which is based on AIChE. 1986

84 J. Loss Prev. Process lnd.,?989, Vol2, April

 Two-phase flow venting from reactor vessels: J. C. Leung and Ii. G. Fisher

10 Leung, .I. C., AIChE Journal 1986, 32 (IO), 1743 Subscripts

11 Lung, .I. C. and Grolmes, M. A., AIChE Journal 1987, 33 (3). I entrylocation inside vent line
524 3 outside vent line or ambient location
12 First, K. E. and Huff, J. E., in ‘Design Charts for Two-Phase sample or average value
Flashing Flow in Emergency Pressure Relief Systems’, AlChE : bomb or container
Spring National Meeting Loss Prevention Symposium Paper No. c critical or choked
73d, New Orleans, 6-10 March, 1988 f liquid phase
13 Grolmes, M. A. and Lung, .I. C., Chew?. Eng. Prog. 1985,81(S), 0 initial condition, inlet stagnation, frictionless nozzle
47 “C choked condition for frictionless nozzle
14 Fauske & Associates, Inc., in ‘Systems Analysis for Integrated
P peak turnaround condition
Relief Evaluation @AFIRE) Users’ Manual’, SAFIRE Computer 5 set pressure condition
Program and Documentation, DIERS/AIChE, 1986
15 Fauske & Associates, Inc., in ‘Phase III Large Scale integral Tests,
DIERS III-S, Experimental Results for Series III Tests’,
Small/Large Scale Experimental Data and Analysis, DlERS/
AIChE, 1986
16 Fauske & Associates, lx., in ‘Phase III Large Scale Integral Data,
DIERS 111-6, Experimental Results for Series IV Tests, Analysis
Appendix
and Program Summary’, Small/Large Scale Experimental Data Given the styrene example in Table 2 and a 200 L/D
and Analyses, DIERS/AIChE, 1986
equivalent relief line with a diameter of 0.35 m, evaluate
17 Shapiro, A. H., in ‘The Dynamics and Thermodynamics of
Compressible Fluid Flow’, The Ronald Press Company, New the overpressure for the following cases and illustrate
York, 1953, Vols. 1 and 2 the discharge flow calculation:
18 Duxbury, H. A. and Wilday. A. .I., in ‘Calculation Methods for
Reactor Relief: A Perspective Based on ICI Experience’, IChemE
Symposium on Hazards from Pressure: Exothermic Reactions,
Case I: horizontal vent line
Unstable Substances, Pressure Relief and Accidental Discharge’. cos 8 = 0
1987, 107, p. 175 Fr=O
C, = GJG,, = 0.7 (Figure 6 with Fr = 0 and w = 25)

Nomenclature (a) PJ = 101,000 Pa

93 = 101,000/450,000 = 0.224
nc = 0.9 C, = 0.63 [Equation (16)]
Flow is choked since 73 < nc and Figure 6 is applicable.
liquid specific heat at constant pressure
critical discharge coe&ient Cr=l
discharge coefficient
correction factor due to unchoked flow
co = 0.7
relief vent diameter A = ~0~14 = 0.0962 m 2
Fanning friction factor
Froude number, Equation (8)
gravitational constant
A,rr/Ao = i&A/ Ao = o.7r;;y) = 0.11
mass flux or nms flow rate per unit area
normalized or dimensionless ma flux APIPS = 0.2 (Figure 8 with p = 0.4)
latent heat of vaporization
gas specific heat ratio (h) PJ = 300,000 Pa
pipe length
molecular weight ~3 = 0.667
mass
initial mass in vessel Flow is now unchoked since q~ > Q
4f L/D, jlow resistance factor
absolute system pressure (1 -r)j)/(I-nc) = (1 -0.667)/(1- 0.63) = 0.9
vessel stagnation pressure
back pressure
Cr=G}G,= 1.0 (Figure 7 with w = 25)
set pressure C,, = C,Cr = (o-7)(1.0) = 0.7
pure component vapour pressure
overpressure, Pp - Ps Same AP/P, as in (a) despite the unchoked flow. Note
heat release rate per unit ma that Equation (10) yields a G/, Pspo of 0.105 and a CD
total volumetric generation rate
time
of 0.6, resulting in a somewhat larger AP/P$ of 0.25.
system temperature
temperature rise above set corresponding to AP Case 2: vertical vent fine
temperature rise for reaction rate doubling
specific volume
cos 0 = 1.0
difference between gas and liquid specific volume p,, = me/V= 9500113.2 = 720 kg m-3
volume of vessel
quality or mass fraction of vapour (720)(9.81)K’.35)(1) =o28

ith component liquid mass fraction

(4)(0.005)(450,ooo) .
ith component vapour mass fraction
inlet void fraction C, = G,/G, = 0.47 (Figure 6 with Fr = 0.28 and
parameter, Equation (24)
phi-factor
w = 25)
v/m normalized specific volume
parameter ratio relative to PO (a) Ps = 101,000 Pa
parameter, Equation (4)
density
~3 = 0.224
pipe inclination angle with vertical qc = (0.9)(0.47) = 0.42 [Equation f16j]

J. Loss Rev. Process lnd., 7989, Vat 2, April 85

Two-phase flow venting from reactor vessels: J. C. Leung and H. G. Fisher

Flow is choked since 13 <Q and Figure 6 is applicable. 0.43, which is very close to the above result. This may be
fortuitous since Figure 7 is only intended for horizontal
(0.47I.yQ) = 0.074
lieff/Ao = CDA/AO = flow (Fr = 0). Equation (10) is recommended for Fr > 0
situations.
API Ps = 0.45 (Figure 8 with p = 0.4)

(b) 4 = 300,000 Pa Discharge jlow calculation at 450 kPa

Withw=25andpo=9500/13.2=720kgm-3
1)3= 0.667 From nozzle chart Figure 3, G* = 0.18
Flow is unchoked since ~3 > qc G,, = G*(Pp,,)‘.’ = 0.18[(450,000)(720)]“~’
(1 - v~)/(l- Q) = (l- 0.667)/(1- 0.42) = 0.57 =3240kgm-*s-l
CF = 0.95 (Figure 7 with w = 25)
For Case l(a) above with Co = 0.7:
co = (0.95)(0.47) = 0.45
C= CoGoc = 2270 kgm-2s-’
A,rr,Ao= (0.45)(0.0962) = Q07
0.613 and the discharge mass flow in the vent line is
A P/ P, = 0.55 (extrapolation from Figure 8) CA = (2270)(0.0962) = 218 kgs-’
Equation (10) yields a C/fiF” of 0.078 and a CD of Other cases can be obtained in a similar fashion.

86 J. Loss Rev. Process Ind., 1989, Vo/ 2, April