Sei sulla pagina 1di 7

Sizing pressure relief valves in flashing and

two-phase service: an alternative


procedure

F. N. Nazario and J. C. Leung*


Exxon Research and Engineering Co., Florham Park, NJ 07932, USA
*Fauske & Associates Inc., Burr Ridge, IL 60.521, USA

In view of the greater accuracy of the w HEM correlation and the resulting conservative PR valve
sizing (relative to the traditional PR valve sizing methods), the w HEM-based flow calculations are
recommended for all situations where flashing or two-phase flow occurs within the PR valve.
However, as with any correlation, the accuracy of the w HEM correlation results drops as the
system to which it is applied diverges from the data used to develop it. Analysis of 15 different
systems representative of actual refinery streams indicate that for fluids with a wide boiling range
and for very non-ideal systems such as those containing hydrogen, the HEM correlation
underpredicts the mass flux significantly (i.e. overpredicts the PR valve size). Since the large
deviations are on the conservative side, the procedure will result in excessively large PR valves
that may cause problems because of chattering of the PR valves, but would not present potential
for vessel failure. The majority of the deviations are a result of the fact that the simplifying
assumptions built into the correlation for single component do not truly characterize the actual
flashing behaviour of many multicomponent fluids. The alternative approach presented, in which
the correlation parameter w is based on the actual flashing behaviour, eliminates nearly all the
deviations and significantly improves the results of the correlation regardless of the system
analysed. For trouble systems (those containing more than 0.1 wt% hydrogen and for some
multicomponent fluids with a nominal boiling range greater than 80°C (15O”F), the alternative
approach should be used to define the correlation parameter w. Use of this alternative approach
is valid for any system (and will improve the accuracy of the correlation) but does require an
additional flash calculation. However, for other systems the original formulation of w is adequate
and can be used.

(Keywords: PR valves; flashing flow; two-phase service; flow calculation)

Pressure relief (PR) valves are the primary means of flow through PR valve nozzles in flashing service is the
vessel overpressure protection in refineries and che- homogeneous equilibrium model (HEM) method-
mical plants. Some of these PR valves must operate ology. The traditional API method’ for sizing PR
with fluids that are either two-phase at the valve inlet or valves in flashing or two-phase service can result in
flash as the fluid moves across the PR valve nozzle. The undersized PR valves. The analysis documented in
most significant aspect of flashing flow through a nozzle Reference 1, comparing the HEM results with the
is the ease with which the maximum flow (or choked limited experimental data available and with the API
flow) condition is attained. While for vapour or gas method, indicates that the potential for undersizing the
flow the choking condition is reached when the PR valves is significant, with the PR valve sizes
downstream is less than 55 to 65% of the upstream determined from the API method being as little as
value (in absolute terms), in the case of saturated one-half of the required size as based on the experi-
liquids and flashing two-phase mixtures, the choking mental data. This is caused by the API method
condition may be attained when the downstream significantly overestimating the critical mass flux
pressure is less than 80 to 90% of the upstream through nozzles, as illustrated (for saturated water) in
pressure. Hence, even a small amount of flashing is Figure I. Since the PR valve relief area is inversely
sufficient to significantly limit the flow through the proportional to the mass flux, overestimating the mass
valve. flux results in undersizing the PR valve. The level of
As discussed in Reference 1, the most soundly undersizing actually present in installed PR valves is
based procedure currently available to determine the dependent on many factors including the sizing con-
tingency for the PR valve, the pressure, temperature
and composition of the system in question, and the
Received 16 March I992 degree of oversizing forced by selecting from the
09504230/92/05026347
@I 1992 Butterworth-Heinemann Ltd

J. Loss Prev. Process Ind., 1992, Vol5, No 5 263


Sizing pressure relief valves: F. N. Nazario and J. C. Leung

commercially available PR valves offered. It is worth I I I I I I


noting that the ASME flow capacity curve for rating
nozzle type safety valves on saturated water3 is in good
agreement with the HEM curve as shown in Figure 1.
The classical HEM procedure is well-known but
was not frequently used because it required a large
amount of thermophysical data and the calculations
were tedious’. A major recent improvement is the
development of a correlation based on an analytical
solution of the HEM equations4. This so-called w
correlation allows the use of the HEM methodology
based on the physical properties of the fluid at the PR
valve upstream pressure. This correlation, originally
developed for single component fluids, has been
utilized for multicomponent systems with limited suc-
cess1a5. However, the many simplifying assumptions
built into the correlation result in conservative results
in situations where the real fluids behave differently.
As revealed in the present study, serious problems are
posed by wide boiling range and hydrogen-containing
fluids. Since the use of the w HEM correlation can lead
to excessive relief area being installed, an alternative
procedure to correct the excessive conservatism is
presented in this paper.
2 4 b 8 I” IL 14
Detailed evaluation requirements Stagnation pressure, PO CMPa)

Extensive physical property data Figure1 Comparison of mass flux: API, ASME and HEM (for
Flow of a flashing two-phase (vapour and liquid) saturated water). Data from Uchida and NariaG (A),Fauskeg
(C). Henry10 (O), Allemann eta/.JJ (+) and API-RP 520’* (0)
mixture or a flashing saturated liquid through a nozzle,
such as those used in PR valves, can be treated much
like the classical compressible fluid that undergoes an
isentropic expansion as the pressure drops through the
nozzle. Since the flashing vapour-liquid mixture be- This procedure for evaluating the flow through
haves as a compressible fluid, more fluid will flow and nozzles has been available in the literature for some
expand through the nozzle as the pressure downstream time but is rarely used for sizing PR valves due in part
of the nozzle is lowered until a maximum value is to the large amount of physical property data and
reached. This maximum is known as the choked or vapour-liquid equilibrium data that is needed.
critical mass flux for the nozzle and the pressure at
which it occurs is known as the choked or critical (flow)
pressure. Critical mass flux correlation based on an
The critical mass flux through the PR valve (W/A) assumed expansion law
can be evaluated from the following expression derived As noted, the major drawback of using the above
from the laws of thermodynamics: methodology is that it requires detailed knowledge of
TI, r rP-O 10.5I the vapour-liquid equilibrium, and the calculations,
2 = (2(/z, - h))0.5/u = [ZIP0 dP] /u
even with the aid of a computerized flash routine, are
therefore tedious and time-consuming. Recently, a
To evaluate this expression, detailed thermodynamic
critical flow correlation* was developed which was
properties for the two-phase mixture are needed. A
based on a number of simplifying assumptions about
numerical procedure to evaluate the integral can be
the expansion behaviour of the two-phase fluid that
used if detailed knowledge of the two-phase specific
permitted the expression for the flashing or two-phase
volume and the associated vapour-liquid equilibrium is
mass flux to be solved analytically and an equation for
available. This can in fact be done by successive flashes
the maximum mass flux derived. These assumptions
of the fluid starting at the nozzle upstream pressure and
included the following:
continuing to successively lower pressures. From these
flashes the behaviour of two-phase specific volume, U, 0 single component system
can be obtained as the downstream pressure, P, is l ideal gas behaviour
lowered. The critical mass flux is given by the max- l the fluid is far from its thermodynamic critical point
imum of this expression (Equation (1)) and the choking (T, s 0.9 or P, S 0.5)
pressure for the fluid is the pressure at which the l the heat of vaporization and the heat capacity of the
maximum mass flux occurs. fluid are constant throughout the nozzle (therefore

264 J. Loss Prev. Process Ind., 1992, Vol5, No 5


Sizing pressure relief valves: F. N. Nazario and J. C. Leung

they can be characterized by their values at the to o) can be used in Equation (1) to obtain an
upstream pressure) analytical solution of the mass flux as a function of the
the behaviour of the fluid vapour pressure with pressure ratio from which an expression of the max-
temperature follows the Clapeyron equation imum mass flux can be derived.
isenthalpic (constant enthalpy) flow process Comparison of the analytical solution to data for
the maximum mass flux of single component fluids
Once these assumptions are made, an analytical expres-
through ideal nozzles indicated that a reasonable
sion for the normalized two-phase specific volume (the
correlation could be obtained but deviations from
ratio of the two-phase specific volume at pressure P
constant enthalpy assumption resulted in noticeable
relative to the initial two-phase specific volume at P,)
errors (- 10%) as the fraction of vapour upstream of
as a function of the pressure ratio (the ratio of the
the nozzle approached 1.0 (i.e. w approached 1.0; see
pressure P to the upstream pressure P,) can be
Figure2). (However, for w > 4, there is excellent
obtained. This expression,
agreement between data and theory.) To reduce this
error, the pure fluid data was correlated by Leung using
($_I 1
V
-_=w +1
V” w as the correlation parameter. As can be seen from
Figure2, this correlation fits the data extremely well
can be characterized as a straight line whose slope
(standard deviation < 1%). These explicit curve-fitted
yields the correlation parameter w. This in turn allows
expressions for the maximum mass flux (corrected for
o to be based solely on the physical properties of the
non-ideal nozzle behaviour) and for the critical (flow)
fluid at the upstream conditions (denoted as w) and can
pressure ratio (the characteristic ratio between the
be represented by the following equation:
nozzle upstream and downstream pressures at which
the critical flow occurs), are as follows:
0=.X 0

nc = 0.55 + 0.217 (In o) - 0.046 (ln w)’


The analytical expression of the normalized specific
volume (described as a straight line with the slope equal + 0.004 (In u)~ (4)

,-

- 15%

,’
,’
,’
_’

0 Propane/octane
0 Butaneletbenzene
A ButadienelZM heptane
V Methane/propane
+ Methanelcyclohexane
X Hydrogen/octane
q Methane/propane/octane
E Propaneloctaneltridecane
e Hzlpentaneletbenzene
m Depropanizer feed
ffl Naphtha H/F Hot separator feed
‘23 Wild naphtha
@ C/F HP Hot separator feed
0 R/F HP Hot separator feed
l R/F HP Hot separator liquid

I I I I I I I I
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 .o

Theoretical tnass flux


Figure 2 Correlation of critical flow parameters

J. Loss Prev. Process lnd., 1992, Vol5, No 5 265


Sizing pressure relief valves: F. N. Nazario and J. C. Leung

were representative of actual multicomponent refinery


for 1 6 w < 4 (5) streams. These results were compared to the maximum
mass flux obtained from the w HEM correlation and
W P are presented in Figure 3. This figure is a parity plot of
- = K&, 0 the correlation results against the theoretical calcu-
A J VO
lation and includes two limiting lines at f15% to
x 0.6055 + 0.1356 (ln w) - 0.0131 (ln w)* indicate the maximum acceptable error in the calcula-
G tion procedure.
for4s ws 100 (6) As shown in Figure3, there are a number of
The correlation for the mass flux is very accurate systems that completely fall outside the acceptable
(-Cl%) relative to the pure fluid data used to develop band while for other systems the results vary depending
it4 and also reasonably accurate (+lO%/-3%) when on the temperature and pressure considered (i.e. on the
compared with the limited data available for the flow of degree of vaporization or quality of the stream at the
close boiling binary mixtures through nozzles’. PR valve inlet). All the large deviations (i.e. the
However, the correlation is sensitive to the physical difference between the prediction from the correlation
property data used and accurate data are needed to and the calculated theoretical maximum mass flux) are
obtain reasonable answers. on the conservative side. Hence, if the HEM corre-
In view of the increased accuracy of the HEM lation is applied to these non-ideal systems, the
correlation and the resulting conservative PR valve procedure will result in excessively large PR valves.
sizing (relative to the traditional PR valve sizing Thus may cause problems because of chattering of the
methods), the HEM-based flow calculations are re- PR valves, but would not present a potential for vessel
commended for all situations where flashing or two- failure.
phase flow occurs within the PR valve. The existing Overall, the HEM correlation appears adequate
procedures for this type of service in API RP-520 and for reasonably narrow boiling systems (those with a
API RP-521 should not be used since they may result in difference in nominal atmospheric pressure boiling
undersized valves. The size of existing flashing or temperature between the lightest and heaviest compo-
two-phase service PR valve installations should be nent of less than 80°C (150°F)) that do not deviate too
checked against the size required if the HEM method- far from ideality. But for wide boiling fluids and for
ology is used to verify the adequacy of the overpressure non-ideal systems such as those containing hydrogen, in
protection. particular, where the vapour produced is a result of
lower gas solubility as the pressure drops, the HEM
correlation underpredicts the mass flux significantly
Excessive conservatism of the HEM
(i.e. overpredicts the PR valve size). The reason for the
correlation
deviations vary, but in all cases the systems with the
As with any correlation, with @ HEM methodology large deviations violate one or more of the assumptions
presented results in excessive errors if the system to used in the development of the correlation.
which it is applied deviates substantially from the data For wide boiling fluids, the deviation appears to be
used to develop the correlation. Hence, the correlation caused by the change in physical properties as the
(Equations (4)-(6)) is only valid for 1.0 G w 6 100. composition of the flashed vapour changes with lower
Use of the correlation for w falling outside this range is pressure. Thus, for these systems the deviations be-
not recommended. come reasonable (less than 15%) when the level of
As indicated, only a limited amount of data is vaporization at the PR valve inlet is high (since further
available for multicomponent systems and the data vaporization does not have as significant an impact on
available are for close boiling binaries (such as propane the vapour physical properties).
and butane). Comparisons against these data (as In the case of hydrogen-containing systems, these
documented in Reference 1) show deviations from the are highly non-ideal and follow Henry’s law rather than
experimental values of under 10% which is generally ideal solution behaviour. In these cases, the correlation
acceptable. However, larger deviations can be ex- tends to overpredict the amount of vapour being
pected when the HEM correlation is applied to other generated as the system pressure decreases. This added
systems, particularly those that violate some of the predicted vaporization is reflected in the reduced
assumptions used to develop the analytical solution and calculated mass flux and a larger PR valve.
subsequently the correlation. Based on this understanding on the reasons for the
To illustrate this impact and to define the practical deviations and using the results of all of the systems
limits of the utility of the correlation, Equation (1) was analysed, it is possible to define regions where the
used together with flash calculations to develop the HEM correlation provides adequate results. The ade-
reference (or theoretical) maximum mass flux through quacy of the results is based on the acceptable error
nozzles for a variety of systems. A total of 15 different level being less than 15% when calculated as the
systems were evaluated at conditions varying from absolute value of the deviation divided by the theo-
saturated liquids to almost all (90 mol%) vapour. retical mass flux. It is clear from the systems analysed
Some of the systems were simple binaries while others that the HEM correlation results in excessive errors for

266 J. Loss Prev. Process lnd., 1992, Vol5, No 5


Sizing pressure relief valves: F. N. Nazario and J. C. Leung

0 Propaneloctane
0 Butanetetbenzene
A ButadienelZM heptane
V Methane/propane
+ Methanelcyclohexane
X Hvdroaenloctane
q M;thanelpropaneloctane
M Propaneloctaneltridecane
8 Hzlpentaneletbenzene
S Depropanirer feed
El Naphtha H/F Hot separator feed
Cl Wild naphtha
C3 CIF HP Hot separator feed
0 R/F HP Hot separator feed
n R/F HP Hot separator liquid

Theoretical mass flux

Figure 3 Parity plot: mass flux correlation versus theoretical maw, flux

essentially any system that contains an appreciable Since even for the most non-ideal systems examined,
amount of hydrogen (those in which the hydrogen the normalized specific volume as a function of the
content is greater than 0.1 wt %). For other systems the pressure ratio can be described reasonably accurately
acceptability of the results can be roughly correlated to with a straight line (up to the critical flow pressure
the boiling range, composition and level of vaporiza- ratio), this relationship can functionally be defined by
tion, with the major factor being the boiling range. the physical properties of the vapour and liquid present
Thus, the results from the HEM correlation are at P, (or if the PR valve inlet fluid is saturated, the
acceptable for systems where the nominal boiling range physical properties of the bubble point vapour and
of the components at atmospheric pressure is less than liquid) and the properties of the vapour and liquid at
or equal to 80°C (150”F), regardless of the level of 90% of the upstream pressure P,. (In calculating the
vaporization in the PR valve inlet. In most other physical properties at 90% of P,, the flash calculation
situations the HEM correlation results will be exces- should be carried out isentropically, but an isenthalpic
sively conservative. flash is adequate.) Hence, the appropriate correlation
parameter w (denoted as 09) to use in the HEM
Alternative procedure to eliminate the correlation is given by the slope of this line which can
observed deviations be expressed as:

Analysis of the comparison between the theoretical "Y


mass flux and the mass flux obtained from the HEM
correlation indicates that the majority of the deviations
wy=9
1 --1
“0 1
(7)

result from the fact that the simplifying assumptions do where vg refers to the specific volume evaluated at 90%
not truly characterize the actual flashing behaviour of of the upstream pressure, P,.
the real fluid. Thus, an alternative approach in which Figure4 presents the results of substituting this
the correlation parameter w is based on the actual equation for Equation (3) to define the correlation
flashing behaviour will eliminate all the deviations. parameter w used in the HEM correlation (Equations

J. Loss Prev. Process Ind., 1992, Vol5, No 5 267


Sizing pressure relief valves: F. N. Nazario and J. C. Leung

1.0 requires an iterative calculation to determine the value


.P of the critical (flow) pressure ratio, rlc from Equation
5 0.9
(8), prior to using this value in the analytical formula-
z 08
tion for W/A (Equation (9)), as follows:
2 *
I
h 0.7
7$ + (w” - 20~) (1 - qJ2 + 2w21n (77,)
-iz + 2oZ(1 - Q) = 0 (8)
y 0.6
'L
W --PO rlc
: 0.5 - = KdKb (9)
E A J 0, 6
': 0.4
F Figure5 illustrates the solution in design chart (thick
2 0.3 line with k = 1.0) and shows a smooth transition
E
between the two regimes where the boundary is the gas
2 0.2
.N
flow condition (o = 1 .O) . The results for other values as
z of k are also shown7.
E 0.1
b
z
0
1 10 100
Situation where excessive conservatism can
XJJgo C,,ToPo utg~
W=---+f-
00 ( I
occur
"0 ",a
Use of the alternative approach is valid for any system
Figure4 Parity plot: mass flux based on alternative procedure
versus theoretical mass flux (and will improve the accuracy of the CLIHEM
correlation), but this approach does require an addi-
tional flash calculation and is not always necessary.
However, for systems in which the HEM correlation
(4)-(6)). As can be seen from this parity plot of the using the o parameter based solely on the PR valve
HEM correlation results against the theoretical max- upstream properties results in excessive deviation, the
imum mass flux, nearly all the deviations have disap- alternative approach to defining the correlation para-
peared. In fact, the use of Equation (7) significantly meter w (Equation (7)) should be used to prevent
improves the results of the correlation regardless of the oversizing the PR valves. These troublesome systems
system under consideration. are those having the following characteristics: those
containing more than 0.1 wt % hydrogen or those with
a nominal boiling range greater than 80°C (150°F).
Hence, for systems that fall into these two categories,
A different relationship for the mass flux is
the alternative approach to defining w is re-
required for u) lower than 1.0
commended. For other systems that do not fall in the
As mentioned earlier, Equations (4) - (6) are not valid two categories described above, the original formu-
for values of w lower than 1.0. In this region there is lation of o (Equation (3)) is adequate and may be used.
nominally no flashing and w is generally characterized
by the inlet void fraction (Y,. (This comes from the first
term on the right-hand side of the Equation (3).) This
region, therefore, represents systems composed of a
non-condensable gas and a non-flashing liquid. In fact,
the detailed calculations for the theoretical mass flux
indicate that some systems, specifically those that
expand as a result of lower hydrogen solubility as the
pressure is lowered, are best characterized by CY,values
in this region (i.e. the implied o based on the slope of
the normalized specific volume versus the pressure
ratio from the flash calculations is less than 1.0). This is
not surprising since these closely approach the non-
flashing systems rather than the flashing systems.
For these situations (where o < l), use of Equ-
ations (4)-(6) derived from flashing systems could
result in overestimating the mass flux. Since this would
result in PR valves that would be smaller than required .Ol 0.1

(i.e. not conservative), its use in this region (w < 1.0) is


not recommended. Instead, the former theoretical 6J= a,

solution can be extended to values of w below 1.0 for


the non-flashing flow regime6. The solution, however, Figure 5 Theoretical mass flux design chart

268 J. Loss Prev. Process Ind., 7992, Vol5, No 5


Sizing pressure relief valves: F. N. Nazario and J. C. Leong

References Difference between vaponr and liquid specific enthalpy at


1 Leung, J. C. and Nazario, F. N. J. Loss Prev. Process Ind. 1990, PR valve inlet (J kg-‘) (If all liquid inlets use the physical
3,253 properties of the bubble point vapour)
2 API Recommended Practice 521, Guide for Pressure Relief and Two-phase specific enthalpy (J kg-‘)
Depressmizing Systems, 3rd Edn., American Petroleum Institute, PR valve upstream specific enthalpy (J kg-l)
1990 Discharge coefficient (fraction)
3 ‘Capacity Conversions for Safety Valves’, ASME Boiler and PR valve back pressure correction factor for vapour
Pressure Vessel Code Aooendix 11. American Societv, of Mecha- (fraction)
nical Engineers, 1986,6iO’ Pressure (N mm2)
Critical flow pressure (N mm2) (abs)
4 Leung, J. C. AIChEJ. 1986,32,1743
PR valve upstream pressure = set pressure + accumulation
5 Leung, J. C. and Fisher, H. G. I. Loss Prev. Process Ind. 1989,2,
78 (Nm-*) (abs)
6 L&mg, J. C. AIChE J. 1990,36,797 PR valve upstream temperature (K)
7 Leung, J. C. Chem. Eng. Prog. 1992,88(2), 70 Two-phase specific volume (m3 kg-‘)
8 Uchida, H. and Nariani, H. in Proc. of 3rd Int. Heat Transfer PR valve upstream liquid specific volume = l/p+
Conf., 1966, Vol. 5 (m3 kg-‘)
9 Fauske, H. K. Chem. Eng. Prog. Symp. Ser. 1965,61,210 PR valve upstream vapour specific volume = l/pp,,
(m3 kg-‘) (If all liquid inlets use the physical properties of
10 Henry, R. E. An Experimental Study of Low-Quality, Steam-
Water Critical Flow at Moderate Pressures’, ANL-7740. Argonne the bubble point va our)
National Laboratory, Argonne, IL, USA, 1970 ugO - of0 (m3 kg- 4)
PRvalve inlet specificvolume = ufa + x,ufgO (m3 kg-‘)
11 Allemann, R. T., McElfresh, A. I., Neuls, A. S., Townsend, W.
C., Wilbum, N. P. and Witherspoon, M. E. ‘Experimental High Specific volume of liquid existing at 90% of P, = l/pp
Enthalpy Water Blowdown from a Simple Vessel Through a (m3 kg-‘)
Bottom Outlet’, NBWL-1411. Battelle Northwest Laboratory, Specific volume of vapour existing at 90% of P, = l/pfl
Richland, Washington, USA, 1970 (m3 kg-l)
ups - uf~ (m3 kg-‘)
12 API Recommended Practice 520, Sizing, Selection and Installa-
Specific volume evaluated at 90% of P, = uf~ + x9vfH
tion of Pressure Relieving Devices in Refineries, Part 1 - Sizing
and Selection, 5th Edn., American Petroleum Institute, 1990 (m3 kg-‘)
PR valve relieving rate (kg s-1)
PR valve inlet vapour mass quality (fraction)
Nomenclature Vapour mass quality evaluated at 90% of PO (fraction)
A PR valve relief area required (m*) Void fraction at PR valve inlet
CPfO PR valve inlet liquid specific heat at constant pressure Liquid density at PR valve inlet (kg mm3)
(J kg-l K-‘) Vapour density at PR valve inlet (kgmm3)

J. Loss Prev. Process lnd., 1992, Vol5, No 5 269

Potrebbero piacerti anche