Safety

Rigorously Size Relief

Valves for Supercritical
Fluids
Previously published methods can be tricky to
apply, and may lead to improperly sized valves.
Here is a stepwise, detailed method that
more-accurately determines the orifice area.

Ryan Ouderkirk,
Fluor Corp.

% of Maximum Value

ideal-gas behavior. Relief venting of supercritical fluids

upercritical fluids exhibit characteristics typihas been studied previously, resulting in guidelines for
cal of both liquids and vapors. Transport
calculating the relief rate (3, 4) and the relief-valve oriproperties, such as viscosity and diffusion
fice area (5, 6, 7). However, if one is not careful in readrate, are closer to those of typical vapors,
ing these guidelines, he or she may size the relief valve
while solvent strength resembles that of typical liquids.
based on ideal-gas simplifications or merely for the
Further, the properties can be strong functions of presmaximum mass relief-rate. The reader can inadvertently
sure and temperature, allowing the solvent strength to be
use the vapor or liquid sizing-equations (for an ideal or
easily manipulated. This led to the use of supercritical
compressible gas), for example. Also, Ref. 4 tells the
fluids in industrial processes such as extraction of cafreader how to find the maximum relief rate, as opposed
feine from coffee with carbon dioxide, and the Residuto the maximum orifice area. The maximum required
um Oil Supercritical Extraction (ROSE) process with either butane or pentane (1). However, these same deviaorifice area may not occur at the maximum mass relief
tions from both ideal gas and incompressible fluid berate or the maximum volumetric relief-rate (Figure 1).
havior present distinct challenges for relief valve sizing.
Here is a step-by-step procedure to rigorously size
Here is a rigorous procedure to calculate the relief
the relief valve for supercritical fluids. In presenting
rate and size the relief valve for supercritical
fluids. The relief rate is modeled over time for
a blocked-in vessel using small increments of
Volumetric Relief Rate
temperature. The relief valve is sized by mod100%
eling mass flux through an isentropic orifice up
to the limit of choked flow. An n-butane fire
Orifice Area
80%
case will illustrate the procedure using LeeKesler properties (2). The method is suited for
60%
Mass Relief Rate
other heat-input cases, as well as other property-determination methods.
40%
Many relief problems are simplified signifi1.05
1.10
1.15
1.20
1.25
cantly by assuming incompressible-fluid or
Reduced Temperature
ideal-gas behavior. A fluid at supercritical conditions can be quite compressible, and its phys Figure 1. The maximum required orifice area may not occur when expected
ical properties may deviate appreciably from
during relief venting.

34

www.cepmagazine.org

August 2002

CEP

this method, we will start by discussing

the fluid conditions leading to supercritical relief. Then, the method will be presented in a simplified form, followed by
a detailed step-by-step example to guide
the user through the process,
while highlighting potential problems
with abnormal cases. Afterwards, we
will cover the basis for the
procedure:
the
required-relief-rate
derivation and the relief-valve orificearea derivation. Also, mention is made
of several alternate procedures for sizing the orifice and of the potential for
supercritical relief to turn into twophase relief flow.
Consider a blocked-in vessel full of
liquid with a vessel-relief pressure
greater than the fluids critical pressure.
If heat is added by fire or other means,
boiling cannot occur. Initially, heat
causes the liquid to expand and the vessel to reach relief conditions. The relief
valve can be sized for this condition
with equations available in Ref. 5. If the
heat continues, both the bulk temperature and the vapor pressure increase.
The liquid may begin to flash in the discharge pipe and in the relief device.
Once the vapor pressure exceeds the
valve backpressure, more-sophisticated
sizing methods are required to account
for two-phase relief. Refer to API 520
Part I (5) for guidance on two-phase relief-valve sizing. Finally, if more heat is
added, the fluid may pass through the
critical temperature and the ensuing relief becomes supercritical. Such reliefvalve sizing is detailed here.
The same general sequence of events
may occur in a vessel partially full of
liquid, except that the vapor space is removed initially. The vapor space may
dissolve into the liquid as the solubility
increases, or may pass through the relief device as the pressure increases.
The pressure increases due to the increasing vapor pressure of the liquid
and the compression of the vapor space
by liquid expansion.

Assemble Relief
Case Information
P, T, T, V,
a, f, t, n
(Note 1)

Select Property
Method and
Assemble Data
Lee-Kesler (Note 2)
Pc, Tc, , Mw
AG

Update Iteration Count

n = n +1

Calculate Heat Input

Q = 21,000fa 0.82
(Note 3)

Calculate Fluid
Physical Properties
V, H, S, V
H

Calculate Temperature
for Next Iteration
Tn+1 = Tn+T
Notes:
1) T may require some trial-and-error.
Often, T may need to be decreased
near the orifice sizing iterations.
2) Lee-Kesler property method selected
for example. Other methods can be used,
but refer to step 2 of the example for a
property method discussion. Required
data depends on selected property method.
3) Heat duty depends on relief case. Listed
equation is for the fire-case example.
Adjust equation for other relief cases or
company specific standards.
4) Mass flux may be limited by choked flow,
rather than relief valve backpressure. For
choked flow, use H and V at the choked
conditions, rather than relief valve
backpressure.
5) Refer to discussion on two-phase relief
flow if the property method predicts
two-phase flow in the relief path.
6) For most cases, this is an acceptable
criterion to stop the iterations, but there
may be cases where the orifice area
begins to increase again and further
iterations are required.

Calculate Required
Relief Rates

V = Q V
H

m = V/V

Calculate Time Since

Onset of Fire [Optional]
Vave = (Vn+1+Vn)/2
m = V/Vave
tn+1 = tn+m H
Q /60

Calculate Isentropic
Nozzle Mass Flux
(Notes 4, 5)
So = Sb

G=
No

Are

m, V, A
Decreasing?
(Note 6)

2(Ho Hb)

Vb

3,955.77

Calculate Orifice Area

m
A=
GKbKcKdKv

Yes
Select Relief Valve
Orifice Size

Add Results to Table

Figure 2. A simplified logic diagram for sizing relief valves for supercritical fluids.
Summary of method
The case considered here is heat
from a fire or other means. The fluid does not boil, since
Without boiling, the temperature will continue to rise until
the relief pressure is above the fluids critical pressure.
the heat input is zero. Since supercritical fluid properties

CEP

August 2002

www.cepmagazine.org

35

Safety

Set @ 800 psig

21% allowable
overpressure

Liquid Full,
To = 815R

n-butane
m, V, H, S
Tc = 765.22R
Pc = 550.57 psia
= 0.2002
Mw = 58.12 lb/lbmol

h = 12 ft

D = 6 ft

Insulation,
f = 0.3

2:1 Elliptical
Heads

Figure 3. Fire-case example for n-butane.

can be strong functions of temperature and pressure, the

vessel may require a dynamic model to account for the rising temperature and a potentially increasing relief rate and
orifice area. This procedure embodies such a dynamic
model for the relief device, and includes the sizing procedure from calculating the relief rate through sizing the orifice. The dynamic model is controlled by the increase in
temperature of the vessel contents. For each temperature,
the relief rate is first calculated based on the fluid physical
properties and the heat input; then the relief valve is sized
by assuming isentropic orifice flow (Figure 2).

Stepwise example
Consider a blocked-in vessel initially full of n-butane at
the conditions shown in Figure 3. Relief is needed due to a
fire of unknown duration.
Step 1: Assemble the relief case information:
P, T, T, V, a, f, t, n.
From Figure 3, the vessel is initially at 815R. The relief-valve set-pressure is 800 psig with a 21% allowable
overpressure for the fire case. In reality, the vessel pressure will vary during relief. First, the added heat is enough
to cause the vessel to reach the relief-valve set-pressure.
Then, the valve opens and partially empties the vessel.
The valve may reseat and possibly chatter or flutter since
it may be slightly oversized. The process repeats until the

36

www.cepmagazine.org

August 2002

CEP

fire is extinguished. With a properly sized valve, the vessel

never reaches the allowable overpressure of 21%. However, for simplification, assume that the vessel is initially at
the allowable overpressure and remains so throughout the
relief case. So, T = 815R; P = 1.21 800 + 14.7 = 982.7
psia.
A temperature change must be selected for the iteration. This requires trial-and-error. With decreasing increments, the relief rate and orifice area should converge to
a solution. Often, a large change can be tried to broadly
establish the range that leads to the maximum orifice
area. Then, a smaller change can be used throughout that
range to accurately find the maximum required orifice
area. For the example, 2R appears to be sufficient:
Thus, T = 2R.
Next, calculate the vessel properties. Volume and surface area for different types of vessel heads can be estimated using Ref. 9. Guidelines for the environmental factor, f,
are found in Ref. 3. To calculate the vessel volume, the
wall thickness is ignored, as is the slight expansion caused
by the increasing wall temperature. To find the wetted-surface area, assume that the vessel is completely engulfed in
flames with all of its walls wetted by the supercritical fluid.
The top head area is included in this calculation, but this
may be conservative. Also, the vessel is assumed to be isolated from the piping system, so pipe surfaces are not included. The insulation is fireproof and an environmental
factor of f = 0.3 is assumed.
2
3
D h+2
D = 395.8 ft 3
4
24

(1)

a = Dh + 2 1.084 D2 = 304.23 ft 2

(2)

V=

Now, initialize the variables for time, t, and the number

of iterations, n. t = 0 and n = 0.
Step 2: Select the property-estimation method and assemble the data:
Pc, Tc, , Mw, AG.
n-butane is a nonpolar, single-component hydrocarbon,
so the Lee-Kesler method (2) should suffice. Lee-Kesler, or
a slightly modified version, is available in most commercial simulators. Since supercritical relief typically occurs in
the region of highest uncertainty for most property-determination methods, the physical properties should be
checked against laboratory data, if possible.
Discussing the proper selection of a method is beyond the scope of this article, but Lee-Kesler offers
two major benefits. One is the integration of the thermal properties (enthalpy and entropy) with the other
physical properties. Some methods require a separate
determination for thermal properties. However, thermodynamic cohesiveness may not exist between the

Nomenclature
References denote sources of data.
a = fire case, total wetted surface area, ft2
A = required relief-valve effective orifice-area, in.2
A, B, C, D, E, F, G = ideal-gas thermal property coefficients (8)
D = vessel outside dia., ft
f = environmental factor from API 521 section 3.15 (3)
G = theoretical relief-valve mass-flux, lb/hin.2
h = vessel tangent-to-tangent height, ft
H = relief- or vessel-fluid enthalpy, Btu
Kb = relief-valve capacity correction factor due to backpressure (5)
Kc = relief-valve combination correction factor for rupture disc (5)
Kd = relief-valve effective coefficient of discharge (5)
Kv = relief-valve viscosity correction factor (5)
m = relief- or vessel-fluid mass, lb
n = number of iterations
Mw = molecular weight, lb/lb-mol
Q = relief-case heat input, Btu/h
Qrev= reversible heat input causing entropy to increase, Btu
P = pressure, psia
R = ideal gas constant, 10.73 psia ft3/lb-molR or
1.986 Btu/lb-molR
S = relief- or vessel-fluid entropy, Btu/R
t = time since onset of fire, min
T = temperature, R
v = velocity of fluid through orifice, ft/s
V = Relief- or vessel-fluid volume, ft3
z = fluid compressibility factor, dimensionless
Symbol
= fluid acentric factor, dimensionless
Subscripts
ave= average
b = at outlet of relief valve orifice
c = critical point
IG = ideal gas
r = reduced or residual
0 = initial value or at inlet of relief valve orifice
Superscripts
0, 1= first- and second-residual property terms
per Lee-Kesler (2)
^ = mass, /lb
= time, /h

two methods. The second benefit is the integration of

liquid and vapor properties into a single correlation.
Some procedures use two distinct correlations for liquid and vapor properties, so the following caution
should be observed:
Although there is no distinguishable transition from supercritical liquid to supercritical vapor, some methods
must make a transition from the liquid to the vapor correlation. This may lead to a discontinuity in the physical
property or in the change of the property with respect to
temperature or pressure.
For this relief-valve-sizing method, the property method

must be continuous for both a physical property and the

change with respect to temperature and pressure.
For Lee-Kesler, the critical properties and ideal-gas
thermal property coefficients are required and can be assembled from publications, such as Ref. 8. To calculate
the residual properties with Lee-Kesler and the ideal-gas
thermal properties, the following n-butane properties are
assembled:.
Pc = 550.57 psia; Tc = 765.22R; = 0.2002; MW =
58.12 lb/lb-mol.
A = 7.228140; B = 0.099687; C 104 = 2.66548;
D 108 = 5.4073; E 1011= 4.29269; F 1015 = 6.69580;
G = 0.345974.
Other pertinent information found in the Data Book
(8) are the details of both the Lee-Kesler and the ideal-gas
thermal-property methods, a useful discussion on critical
properties, and recommendations for calculating critical
properties of mixtures.
Step 3: Update the iteration count. For the first step, n =
n + 1 = 0 + 1 = 1.
Step 4: Calculate the heat input, Q. For this fire case, the
heat absorption rate for the vessel is calculated by assuming that Eq. 3 from API 521, Section 3.15.2 (3) represents
the heat absorbed by a supercritical fluid with adequate
drainage and prompt firefighting (4):
Q = 21, 000 fa 0.82 = 21, 000 0.3 304.230.82
= 684, 800 Btu / h

(3)

This equation can be modified for company-specific

standards or other heat-input cases. Ref. 4 suggests that the
heat absorbed by the fluid can more-accurately be modeled
by subtracting out the heat absorbed by the vessel wall. For
simplification, this has been ignored, even though the wall
may act as a significant heat sink.
In this case, the heat duty is assumed to be constant
throughout the relief case. However, for cases such as a
blocked-in heater, the heat duty may vary with vessel temperature, and the duty may need to be recalculated for
each iteration.
Step 5: Calculate the physical properties of the fluid:

V , H , S, V H .
Any of the commercial simulators may significantly
simplify these calculations. For this example, the LeeKesler property method was selected in Step 2. For
more detailed information on Lee-Kesler see Ref. 2.
The fluid is defined by the pressure from Step 1 and the
temperature from either Step 1 for the initial case (n =
1) or the value calculated in Step 12 for all other cases.
These example calculations apply to the initial case, T
= 815R.
If a simulator is used, it will complete most of the remaining calculations in this step. To find the Lee-Kesler

CEP

August 2002 www.cepmagazine.org

37

Safety

properties without a simulator, first compute the reduced

pressure and temperature using the critical properties from
Step 2:
P
982.7 psia
Pr =
=
= 1.785
Pc 550.57 psia
Tr =

3
4
5
S IG = B ln(T ) + 2CT + DT 2 + ET 3 + FT 4 + G
2
3
4
= 1.47525 Btu / lb R

(10)

( 4)
Calculate the actual specific volume, enthalpy and entropy. The molecular weight is taken from Step 2:

T
815R
=
= 1.065
Tc 765.22R

(5)

Next, determine the appropriate residual properties

based on reduced temperature, reduced pressure, and the
acentric factor from Step 2.
z = z 0 + z1 = 0.341304 + 0.2002 0.016560
= 0.344619

(0.344619)(10.73 psia ft 3 / lbmol R)(815R)

(982.7 psia)(58.12 lb / lbmol)
= 0.0527655 ft 3 / lb

(11)

(6)
H RT
H = H IG r c
RTc Mw

H r
H 0
H 1
= r + r = 3.14750 + 0.2002 1.72587
RTc RTc
RTc
= 3.49302

zRT
V =
=
PMw

(12)

(7)
= 278.2615 Btu / lb
H

Sr Sr 0
S 1
P
=
+ r + ln
= 2.3214 + 0.2002 1.6658
14.7
R
R
R
982.7
+ ln
= 6.8573
14.7

H IG = A + BT + CT 2 + DT 3 + ET 4 + FT 5
= 278.2615 Btu / lb

www.cepmagazine.org

August 2002

(13)

(8)

To calculate the residual properties, Lee-Kesler provides both tables and figures for hand calculations and
equations for a computer model. To find the relief rates,
a small temperature change is used to estimate the
changes in the physical properties. Since these changes
can be small, a considerable number of digits need to be
carried through the calculations; a hand calculation is not
recommended. A spreadsheet model was setup with
macros to solve both the Lee-Kesler properties and the
relief-valve orifice area. If a hand calculation is necessary, the enthalpy and specific-volume changes found
later in this step should probably have a minimum of
three significant digits. This means that more digits
should be carried through the procedure when using
smaller temperature increments.
Now, find the ideal-gas thermal properties for the enthalpy and entropy based on Ref. 8 and the coefficients
from Step 2: For the T = 815R initial case:

38

(3.49302)(1.986 Btu / lbmol - R)(765.22R)

58.12 lb / lbmol
= 186.926 Btu / lb

(9)

CEP

S R
1.47525 Btu
=
S = S IG r
lb R
R
M
w

(6.8573)(1.986 Btu / lbmol R) 1.2409 Btu

=
58.12 lb / lbmol
lb R

(14)

Finally, determine the volumetric expansion term by

calculating the specific volume and enthalpy at the relief
pressure, and after the temperature increase selected in
Step 1 (2R), i.e., 982.7 psia and 817R for this first iteration. The expansion term can then be calculated from the
change in properties, but may need to be found outside of a
simulator if using one:
= 189.196 Btu / lb
V = 0.0536716 ft 3 / lb; H
V 0.0536716 ft 3 / lb 0.0527655 ft 3 / lb
=

189.196 Btu / lb 186.926 Btu / lb

H
= 0.000399163 ft 3 / Btu

(15)

Step 6: Calculate the required relief rates: V , m .

The derivations of the relief-rate formulas are presented
later on in this article. The required relief rates for the first
iteration are found using the heat input and physical properties from Steps 4 and 5:

V
V = Q
= 684, 800 Btu / h 0.000399163 ft 3 / Btu
H
= 273.3 ft 3 / h
(16)

flux. Find the choked mass flux by maximizing the mass

flux:

G=
273.3 ft 3 0.0527655 ft 3
m = V / V =
/
= 5,180 lb / h
h
lb

2 H 0 H b
Vb

) 3, 955.77

(21)

(17)
subject to the following isentropic constraint:

Step 7: Calculate the time elapsed since the onset of the

S0 = Sb
(22)
fire, t (this step is optional).
This time interval is not needed to find
Table 1. n-butane mass flux for the example.
the required orifice area. However, the
computation may prove useful for fires of
Vb, ft3/lb
Hb, Btu/lb
Tb, R
Pb, psia
Sb, Btu/lbR G, lb/hin.2
limited duration, or when calculating ves815.0
982.7
0.05277
186.93
1.2409
0
sel-wall temperature vs. time. The deriva813.8
963.0
0.05329
186.73
1.2409
46,620
tion is presented later on along with a
783.2
642.9
0.07135
183.15
1.2409
152,310
method for estimating the vessel-wall tem781.2
630.1
0.07287
182.99
1.2409
152,350
perature. The time for the first iteration (n
779.1
617.5
0.07449
182.81
1.2409
152,390
777.0
605.1
0.07621
182.63
1.2409
152,110
= 1) is 0, so, in this example, the calcula774.7
593.0
0.07807
182.46
1.2409
151,440
tion actually applies to the second iteration (n = 2).
First, estimate the average specific volume during the
The pressure at which this maximum flux occurs is the
temperature increase using the specific volumes from Step
choking pressure. The pressure at point b cannot be less
5:
than the choking pressure, even if the relief valve backpressure is less than the choking pressure.
3
3
0.0527655 ft / lb + 0.0536716 ft / lb
(Vn+1 + Vn )
To find the choke pressure, first assume that the ori
Vave =
=
fice inlet properties are equivalent to the vessel proper2
2
ties. Ignore choked flow by assuming that the orifice
= 0.0532186 ft 3 / lb
(18)
outlet-pressure is equal to the relief-valve backpresNext, using the vessel volume from Step 1, find the avsure. Next, calculate the mass flux for multiple backerage vessel mass during the temperature increase. Be
pressures by keeping the entropy constant and starting
aware that mass is removed by the relief device, so the veswith a backpressure that is equal to the relief-valve
sel mass should be recalculated for each iteration:
inlet pressure. A sample calculation for the first iteration for n-butane is summarized in Table 1, with each
= 395.8 ft 3 / 0.0532186 ft 3 / lb = 7, 437 lb (19)
backpressure representing a 2% decrease in the gage
m=V/V
ave
pressure. Once the maximum mass flux is found, the
Finally, calculate the time required for the temperature
choked condition has been solved and no further calcuto increase and add it to the calculated time from the previlations are required. When generating these mass flux
ous iteration. This particular calculation is for the second
values, the property package should be set up to indiiteration, and the time from the previous iteration is zero.
cate two-phase flow of the fluid being relieved.
This calculated time is the cumulative time since the onset
To calculate physical properties for Table 1, refer to
of the fire:
Step 5. The following sample calculation shows the
mass flux calculation for the choked flow condition of

H
Table 1:
t n +1 = t n + m
= 0 + 7, 437 lb
Q 60
189.196 Btu / lb 186.926 Btu / lb
= 1.5 min (20)
684, 800 Btu / h 1 h / 60 min
2 (186.926 182.809) Btu / lb
G=
0.074487 ft 3 / lb
Step 8: Calculate the isentropic-nozzle mass flux, G.
lb
ft 3
152, 390 lb
For more information, refer to the orifice-sizing derivation
(23
=
3, 955.77 2
in. h Btu lb
h in.2
presented later on. Due to choked (or critical) flow, an iterative method may be necessary to find the theoretical mass

CEP

August 2002 www.cepmagazine.org

39

Safety

Table 2. n-butane relief-rate for the example.

H, Btu/lb
S, Btu/lbR
V, ft3/h
m, lb/h

t, min

T, R

V, ft3/lb

0.0
1.5

815
817

0.05277
0.05367

186.93
189.20

1.2409
1.2437

273.3
282.3

9.9
11.2
12.5
13.7
15.0

829
831
833
835
837

0.05976
0.06086
0.06198
0.06312
0.06427

202.93
205.21
207.47
209.73
211.98

1.2604
1.2631
1.2659
1.2686
1.2713

23.7
24.7
25.6
26.6
27.5

853
855
857
859
861

0.07375
0.07495
0.07614
0.07733
0.07852

229.35
231.45
233.52
235.58
237.61

33.3
34.0
34.8
35.5
36.2

875
877
879
881
883

0.08667
0.08781
0.08894
0.09006
0.09117

251.40
253.31
255.20
257.07
258.94

G, lb/hin.2

A, in.2

5,180
5,261

152,390
149,320

0.03486
0.03613

331.1
338.1
344.7
351.0
356.9

5,541
5,555
5,562
5,561
5,554

132,740
130,450
128,180
126,110
124,210

0.04282
0.04368
0.04450
0.04523
0.04586

1.2918
1.2943
1.2967
1.2991
1.3015

391.5
394.3
396.9
399.1
401.1

5,309
5,262
5,212
5,161
5,109

111,950
110,800
109,680
108,680
107,730

0.04864
0.04870
0.04874
0.04871
0.04864

1.3174
1.3195
1.3217
1.3238
1.3259

408.2
408.4
408.4
408.3
408.0

4,710
4,651
4,592
4,533
4,475

102,230
101,600
100,990
100,410
99,830

0.04725
0.04695
0.04664
0.04631
0.04597

From the table, choked flow occurs near 779.1R and

617.5 psia. If the relief valve backpressure is greater than
this choking pressure, then the theoretical mass flux for
the relief valve is equal to the calculated mass flux at that
backpressure. If the backpressure is less than this choking
pressure, then the relief valve is choked and the theoretical mass flux is equal to the calculated mass flux at the
choked condition. Flow will normally be choked, since
the choking pressure is typically greater than 50% of the
relief-valve set pressure (gage), while the relief-valve
backpressure is typically less than 10% of the set pressure
for conventional valves, and 3050% for bellows valves
(5). In this example, the backpressure is assumed to be
less than 10% of the set pressure (94.7 psia). This is well
below the choking pressure (617.5 psia), so that the flow
is choked and the theoretical mass flux is approximately
152,390 lb/hin.2
Step 9: Calculate the orifice area, A.
More on the orifice area and the effective coefficient of
discharge is presented at the end of this article. Based on
the theoretical mass flux and the required relief rate, the required orifice area can finally be calculated. To correct this
calculation for system nonidealities, use the derating factors from Ref. 5 for the backpressure, use of a rupture disc,
discharge coefficient, and viscosity. A conventional valve is
chosen with a backpressure of less than 10% of set pressure, so Kb = 1.0. A rupture disc is not used in the system,
therefore, Kc = 1.0. The effective coefficient of discharge
for the vapor is assumed, yielding Kd = 0.975. Normally,
no viscosity correction is required for a supercritical fluid,
thus, Kv = 1.0.
From Steps 6 and 8, the required relief rate for the initial case is 5,180 lb/h and the theoretical mass flux is
152,390 lb/hin.2 The area is then:

40

www.cepmagazine.org

August 2002

CEP

A=
=

m
G Kb Kc K d Kv
5,180 lb / h
2
152, 390 lb / h in. 1 1 0.975 1

2
= 0.03486 in. (24)

Step 10: Tabulate the results. Summarize the calculations for each temperature iteration by adding a row to a
results table. Once all iterations are complete (refer to Step
11), the table will take the form of Table 2. This table is
truncated to show the first two calculation steps, plus the
steps leading to the largest mass relief-rate, the largest required orifice-area, and the largest volumetric relief-rate.
Step 11: Are m , V and A decreasing? Determine if the
procedure is complete. In most cases, the relief-valve sizing
is complete when the required mass and volumetric reliefrates, as well as the required orifice area, all decrease from
the previous iteration. However, further iterations may be
needed, since the orifice area may begin to increase again or
if the vessel-wall temperature is being modeled for a specific time interval. The orifice area may increase again if the
fluid is a liquid and has not reached supercritical conditions,
since a supercritical fluid typically has a greater expansion
rate than does a liquid. Not all of these cases can be identified, and some engineering judgment may be required to
determine if the relief-valve sizing is complete. The criteria
presented here serve only as guidelines.
If the iterations are complete, skip to Step 13. If additional iterations are required, proceed to Step 12.
Step 12: Calculate the temperature for the next iteration, T.
This is required when further iterations are needed to
complete the dynamic relief-valve model. The temperature must be increased for the next iteration of vesselfluid temperature since boiling does not occur to limit

the temperature, which will continue to rise as long as

heat is added. The calculation shown here is the last step
for the first iteration of the n-butane example; Tn+1 = Tn +
T = 815R + 2R = 817R. This temperature is not the
vessel-wall temperature. Refer to the relief-rate derivation presented later on if an estimated wall temperature
is required. Once the temperature is calculated for the
next iteration, return to Step 3.
Step 13: Select the relief-valve orifice size. Proceed to this
step only when all of the iterations are complete. The required relief valve area is the maximum required orifice area,
A, from the table generated in Step 10. For the example, the
maximum required orifice area from Table 2 is 0.04874 in.2
and a D orifice per API 526 (10) is therefore selected.

Required relief-rate derivation

The basis is presented here for calculating the required relief-rate for heat input with no boiling present.
The method is similar to that summarized in Ref. 4 and
is based on the same main assumptions: there are no
temperature gradients in the vessel and mass does not
exit the system except through the relief device. The
first assumption is conservative with respect to the required orifice area, since the worst-case expansion temperature is assumed throughout the vessel. In reality,
heat is transferred by natural convection of the vessel
contents, during which heat is removed from the wall.
This takes place more slowly than heat-removal boiling,
and can lead to thermal gradients across the contents, an
excessive wall temperature and eventual vessel failure.
The vessel wall may require additional temperature protection as discussed in API 521, Section 3.15 (3). If necessary, approximate the vessel-wall temperature by assuming that heat transfer occurs predominantly by natural convection from the wall to the contents (4).
The required relief rate is derived using the definition of
specific volume:
V = mV

(25)

Evaluating the derivative as a function of time shows the

volumetric growth of the fluid that is absorbing the heat. This
volume increase must be vented since the vessels volume is
virtually constant. Although mass leaves the vessel, the mass
of the fluid expanding is treated as constant. Additionally, the
vessel relief-pressure is assumed to be constant. Thus:
dV
dV
=m
Required relief rate = V =
(26)
dt
dt
The volume increases due to the heat input from the fire
or other relief case. The energy balance for the expanding
fluid is shown below. Again, the relief pressure and fluid
mass are treated as constants.

dH
(27)
dt
Combining Eqs. 26 and 27 eliminates mass and time. The
physical properties are taken across the constant relief pressure:
Q=m

dV
V = Q
dH P

(28)

The differential expansion term is approximated to

simplify calculations:
V
V = Q
(29)
H
The mass relief-rate is then related to the volumetric relief-rate by the specific volume:
m = V / V

(30)

Eqs. 29 and 30 form the basis for Step 6. Eq. 28 is similar to Eq. 1 in API 521, Section 3.14.3 (3), which is used to
describe the thermal expansion of a liquid. However, the
specific volume and enthalpy are typically easier to obtain
from property packages. Eq. 29 would also be applicable to
a boiling-liquid if using the saturated properties and taking
credit for the liquid volume removed by boiling. However,
this credit is typically ignored and the mass relief-rate for a
boiling liquid reduces to the more familiar equation of the
heat input divided by the latent heat (3).
For a boiling liquid, the specific volume and enthalpy
changes are distinct steps in the transition from saturated liquid to saturated vapor. For single-component systems, these
steps occur with no temperature change, mandating that the
specific heat and cubical expansion are infinite for the phase
transition. Theoretically, a similar phenomenon may occur at
the critical point, except with no visible phase change.
The heating time for Step 7 is derived from Eq. 27 by approximating the differential, solving for time, and adding
unit consistency:
tn+1 = tn + m

H
Q 60 min / h

(31)

Relief orifice-area basis

The theoretical basis for calculating the required orifice area
is presented here. The theory of compressible flow through an
orifice or nozzle is well-documented, based upon adiabatic,
isentropic flow, combined with a simplified energy balance.
Thermodynamics textbooks cover the theory in more depth
(11). However, most practical engineering applications rely
upon an ideal-gas simplification to calculate choked-flow behavior. Nonetheless, supercritical relief deviates significantly
from such behavior. Before proceeding with finding the orifice
area, we shall review the compressible flow of real and ideal
gases through an orifice. Additionally, since the flow is com-

CEP

August 2002 www.cepmagazine.org

41

Safety

pressible, choked flow will be reviewed. The starting point is

the entropy and energy balance equations that describe basic
nozzle flow.
Minimal heat is added or removed from the nozzle and the
flow is nearly frictionless. This means that the flow is essentially adiabatic and reversible, which equates to isentropic flow:
dQ
dS = rev = 0
(32)
T
The fluid energy balance for adiabatic, frictionless nozzle flow is also well-known:
dH + vdv = 0
(33)
The velocity term from Eq. 33 is a measure of the kinetic energy, while the enthalpy term combines the internal
energy with the pressure-volume energy. Ignoring choked
flow for the moment, the pressure drop across the orifice
lowers the fluids enthalpy and increases its velocity. Integrating Eq. 33 and adding unit consistency results in:

H
25, 037 lb - ft 2 / s 2 / Btu (34)
v b = vo + 2 H
o
b
Since no mass flows through the orifice until the enthalpy
decreases, the initial velocity is zero and the initial enthalpy
is that upstream from the orifice, in this case the vessel.
Eq. 34 is valid until the sonic velocity is reached at the
choked condition. The terms choked and critical are
synonymous, but choked is used here to avoid confusion
with the term supercritical. Choked flow occurs when an
additional pressure drop no longer increases the mass flux.
The mass flux through the orifice is simply the flow velocity divided by the specific volume, as shown by Eq. 35
(with added unit consistency):
vb
1ft 2
3, 600 s

(35)
2

1h
Vb 144 in.
Combining Eqs. 34 and 35 yields Eq. 36, the simplified
form used here:
G=

G=

2 H
0
b
Vb

) 3, 955.77

lb
in.2 h

ft 3
Btu lb

Also, as in Eq. 32, the flow is isentropic:

S = S
0

(36)

(37)

Eq. 36, subject to the constraints of Eq. 37 and chokedflow theory, is the basis for the relief-valve orifice-flow in
Step 8. As mentioned before, choked flow occurs when a
further pressure drop no longer increases the mass flux.
Using Eq. 35, an additional pressure drop simply begins to
increase the specific volume at a faster rate than the velocity increases and the calculated mass flux reaches a maximum. The choked condition is found by calculating the

42

www.cepmagazine.org

August 2002

CEP

mass flux as a function of multiple isentropic relief-valve

backpressures, or by taking the derivative of the mass flux
with respect to pressure at constant entropy, setting it at
zero, and solving the expression.
For this supercritical relief method, the first method is
selected so the mass flux is calculated for multiple pressures, and the choked condition is found at the maximum
mass flux. If the relief-valve backpressure is greater than
the choking pressure, then Eq. 36 is valid for flow through
the orifice. If the relief-valve backpressure is less than the
choking pressure, then the flow is choked and the mass
flux equals the choked mass flux. Refer to the example in
Step 8.
The mass flux found by this method is based on a theoretical model that should be corrected for system non-idealities.
In a manner similar to the Leung omega method (5, 7), the
derating factors from API 520, Part I (5) are assumed to
apply to supercritical relief. Eq. 38 becomes the basis for calculating the required orifice area for Step 9 of the procedure:
m
A =
(38)
G Kb Kc K d Kv
As stated in API 520 Part I, there are no recognized procedures for certifying the capacity of pressure-relief valves
in two-phase service. This is suspected to be true for supercritical fluids. These fluids have the characteristics of both
liquids and vapors, which adds further complexity. Since
compressible flow theory is used to calculate the mass flux
for the example, the effective coefficient of discharge for
the vapor (Kd = 0.975) is assumed. The accuracy of this assumption is currently unknown. See API 520, Part I, Section 3.2 for additional details on Kd.
This method for determining the orifice area is tedious and
leads to numerous calculations. For this reason, some alternative orifice-sizing methods follow. The discussion includes
some basic observations about the methods relative to supercritical relief. These methods typically apply to orifice sizing
(Steps 8 and 9) and may not include relief-rate methods.

Alternative orifice-sizing methods

Four alternatives are examined for calculating the orifice area. They typically apply to the orifice-sizing steps
and may not include relief-rate methods. The first is sizing
per API 520, Part I, Section 3.6 (5). This standard discusses
the ideal-gas limitation to these equations in its Appendix
B. This limitation sufficiently describes a non-ideal, supercritical fluid, so the equations from Section 3.6 are not recommended for a supercritical fluid unless the modifications
from Appendix B are used. This appendix presents a simplified relief-valve-sizing method for real gases based on
equation-of-state thermodynamics along the isentropic relief path. The procedure presented here uses a similar thermodynamic and relief-path basis, except with a more-rigorous model.
Another API-recommended method is found in API

521, Section 3.15.2.1.2 (3). This one is based on ideal-gas

properties, the physical properties of air, no change in fluid
temperature, and an uninsulated vessel with no mass. Section 3.15.2.2 then suggests that a more-rigorous method,
such as in Ref. 4, may be required when these assumptions
are not appropriate. This procedure presented here is similar to that in Ref. 4, but with the added benefit of built-in
orifice sizing.
A third procedure is the Leung omega method (7). The
suggestion is made that the omega factor from Ref. 7 can be
calculated for a supercritical fluid with an isentropic (or
isenthalpic) depressurization equal to 70% of the relief pressure. Omega characterizes the fluids expansion relative to
an ideal, isothermal gas and is usually greater than unity for
a supercritical fluid, since depressurization typically increases the compressibility factor. Using this method may be
valid when omega is relatively constant, which essentially
amounts to a linear approximation of the compressibility
factor as a function of pressure. The method requires only
one isentropic calculation, or even an isenthalpic calculation,
and is simpler than the procedure presented here. A check
may be needed to recalculate omega at the predicted chok-

Literature Cited
1. Perry, R. H., and Don W. Green, eds., Perrys Chemical Engineers Handbook, 7th ed., McGraw-Hill, New York, pp. 22-14 to
22-19 (1997).
2. Lee, B. I., and M. G. Kesler, A Generalized Thermodynamic Correlation Based on Three-Parameter Corresponding States, AIChE J.,
21 (3), pp. 510527 (May 1975).
3. Guide for Pressure-Relieving and Depressurizing Systems, API
Recommended Practice 521, 4th ed., American Petroleum Institute,
Washington, D.C. (Mar. 1997).
4. Francis, J. O., and W. E. Shackelton, A Calculation of Relieving
Requirements in the Critical Region, in Proc. Refining Dept., 64,
American Petroleum Institute, Washington, D.C., pp. 179-182
(1985).
5. Sizing, Selection, and Installation of Pressure-Relieving Devices in
Refineries, Part I - Sizing and Selection, 7th ed., API Recommended
Practice 520, American Petroleum Institute, Washington, D.C. (Jan.
2000).
6. Leung, J. C., and M. Epstein, A Generalized Critical Flow Model
for Nonideal Gases, AIChE J., 34 (9), pp. 15681572 (Sept. 1988).
7. Leung, J. C., Easily Size Relief Devices and Piping for Two-Phase
Flow, Chem. Eng. Progress, 92 (12), pp. 2850 (Dec. 1996).
8. API Technical Data Book Petroleum Refining, API Refining
Dept., Chapters 1 (1997), 4 (1987), 5 (1994), 6 (1984), 7 (1992),
American Petroleum Institute, Washington D.C.
9. Moss, D. R., Pressure Vessel Design Manual: Illustrated Procedures
for Solving Every Major Pressure Vessel Design Problem, p. 224,
Gulf Publishing, Houston, TX (1989).
10.Flanged Steel Pressure Relief Valves, 4th ed., API Standard 526,
American Petroleum Institute, Washington, D.C. (June 1995).
11. Smith, J. M., and H. C. Van Ness, Introduction to Chemical Engineering Thermodynamics, 4th ed., pp. 220225, McGraw-Hill, New
York (1987).

ing pressure, rather than at 70% of relief pressure, to evaluate the extent of change in omega, mass flux and orifice area.
This would require a second isentropic calculation.
The last method for calculating the area is the use of the
generalized charts from Ref. 6. These charts show that the
mass flow of a supercritical fluid can be significantly higher than for an ideal gas (Z = 1). This suggests that using the
ideal-gas formulas in API 520, Part I (5) should normally
yield a conservative orifice area at relatively low reduced
pressure and temperature, when using a compressibility
factor of 1 (Z = 1). These charts from Ref. 6 may lead to
mass fluxes similar to the ones calculated here, since both
depend upon generalized properties and isentropic-nozzle
flow. With additional calculations, the entire relief-valve
sizing procedure for the fire-case orifice-area may perhaps
be generalized. However, the charts may be limited to
cases with constant heat input.

Two-phase relief flow

An interesting two-phase phenomenon can occur during
supercritical relief. As the pressure decreases isentropically, the relief path can pass into the two-phase region of the
fluid. During the mass-flux calculation, the selected property method should be set up to indicate if the fluid has entered the two-phase region. If so, the superheated vapor has
depressurized into a saturated vapor, and may eventually
precipitate some saturated liquid. The saturated liquid may
flash off when depressurized further, allowing the vapor to
return to the superheated state. This can occur because the
saturated vapor entropy and enthalpy typically decline
when approaching the critical point. This can be seen in
numerous entropy-pressure diagrams and tables. To use
this procedure, the physical properties of the two distinct
phases must be combined into a single value, as required
for calculations. One method of doing this is the homogeneous equilibrium model (HEM), which assumes equilibrium of the two phases with no slip flow, and is the basis of
the Leung omega method. A discussion of HEM is found
in Ref. 7. Leungs method is included in the 7th ed. of API
520, Part I (5) as a potential two-phase sizing scheme.
However, the reader is cautioned that two-phase reliefvalve sizing is still evolving and current recommended
practices might be replaced with other ones. Always refer
CEP
to latest API guidelines.

RYAN OUDERKIRK is a process engineer with the Energy & Chemicals unit of
Fluor Corp. (100 Fluor Daniel Drive, Greenville, SC 29607; Phone: (864) 2818656; Fax: (864) 676-7211; Email: [email protected]). His projects
have included conceptual, front-end, detail and plant engineering design, as
well as commissioning and start-up of a process unit based on new
technology. His design experiences have included viscous fluids, fluids near
their critical point, and a pilot-plant scale-up. He received his bachelors
degree from the Georgia Institute of Technology. Thank you to my coworkers
at Fluor who helped with editing and content, and a very special appreciation
to my father, Andrew.

CEP

August 2002 www.cepmagazine.org

43