Vous êtes sur la page 1sur 11

Journal of Loss Prevention in the Process Industries 27 (2014) 21e31

Contents lists available at ScienceDirect

Journal of Loss Prevention in the Process Industries


journal homepage: www.elsevier.com/locate/jlp

API Standard 521 new alternative method to evaluate re relief for


pressure relief device sizing and depressuring system design
Edward Zamejc
EZ Relief Systems Consulting, Inc., 7905 Grady Circle, Castle Rock, CO 80108-6112, USA

a r t i c l e i n f o a b s t r a c t

Article history: Since the 1950s, API Standards have provided guidance on determining relief loads for equipment
Received 31 October 2012 exposed to pool res. The API method is empirical based on tests performed in the 1940s. There is
Received in revised form increasingly widespread interest in analytical methods based on heat transfer principles to model re
28 October 2013
heat input. The API committee agreed to include an analytical method in the 6th edition of API Standard
Accepted 29 October 2013
521 to establish relief loads for pressure relief devices and to design depressuring systems for the re
scenario. The analytical method provides more exibility than the empirical method but has limitations
Keywords:
(e.g., too many permutations are possible leading to potential under-sizing of the pressure relief device).
Pool re exposure
Jet re exposure
This paper discusses the basis for the empirical method in API Standard 521 and provides comparisons
Vessel failure of the empirical and analytical method with two more recent large-scale pool re tests. This pool re test
Tensile strength data indicates that the empirical method will provide a conservative estimate of pool re heat input for
Time to failure most applications and is still the method of choice when designing pressure relief systems. However,
Fire heat input these recent tests indicate the empirical method needs to be modied when a vessel or equipment is
partially conned by adjacent embankments or walls equal or greater than the vessel height. In such
cases, the wetted area exponent should be 1.0 instead of 0.82.
The analytical method is useful in determining time-versus-temperature proles for heating unwetted
vessels of varying wall thicknesses and materials of construction. These proles, which depend upon the
type of re (e.g., unconned pool re, jet re, etc.), can be combined with tensile strength and stress-
rupture data to specify a depressuring systems pressure-versus-time prole. This will minimize fail-
ure and/or mitigate the effects of failure due to overheating from re exposure.
2013 Elsevier Ltd. All rights reserved.

1. Introduction depressuring systems for the re scenario. The analytical method


will complement, but not replace, the existing empirical method.
American Petroleum Institute (API) Standard 521 Pressure It is important to establish the scope of API 521 and differentiate
Relieving and Depressuring Systems is an internationally recog- it from API 2000. API 521 covers pressure vessels and processing
nized engineering standard used to design pressure relief systems, equipment (e.g., vessels design in accordance with ASME Section
disposal systems (e.g., ares), and depressuring systems (ANSI/API VIII, Division 1 and similar pressure vessel design codes). In
Standard 521, 2013). It is continually being reviewed, with new contrast, API 2000 covers low pressure, atmospheric, and refriger-
editions published in about 5-year intervals. A technical committee ated storage tanks designed in accordance with storage tank
consisting of industry representatives, engineering contractors, and standards such as API 650. The current pool re heat input equa-
regulators recommend and integrate modications into the Stan- tions in API 2000 are the same as those in NFPA 30. They were
dard. These modications involve lessons-learned from incidents established in a 1963 meeting between API and NFPA and are based
or near-misses, advances in engineering methodologies, and new on a re test and experience with storage tank res. Because the
guidance based on shared experiences of the members or inspired origin/basis of API 521 and API 2000 re equations are different and
by technical inquiries. the scope of the equipment design codes are different, the re
The new API 521 6th edition includes an analytical method to exposure guidance API 521 and API 2000 can neither be inter-
establish relief loads for pressure relief devices and to design changed nor compared (i.e., use API 521 for pressure vessels and
API 2000 for storage tanks). The subsequent discussion relates only
E-mail address: edzamejc@ezreliefsystems.com. to API 521.

0950-4230/$ e see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jlp.2013.10.016
22 E. Zamejc / Journal of Loss Prevention in the Process Industries 27 (2014) 21e31

2. Fire scenario e API empirical method

2.1. Basis of the API empirical method to evaluate pool res

The re scenario generates the most technical inquiries of any


topic in API 521. The current method, given in Equations (1) and (2),
is an empirical method based on re tests performed in the 1940s.
Pool re heat input with adequate drainage and prompt
reghting:

Q C1 $F$Aws 0:82 (1)


Pool re heat input without adequate drainage and prompt
reghting:
Fig. 1. Pool re heat input versus wetted area exposure.
Test data sources: a) API project test No. 1, b) API project test No. 2, c) Rubber Reserve
0:82
Q C2 $F$Aws (2) Corporation test No. 17, d) Standard Oil Company of California, e) Underwriters Lab-
oratories, Inc. (ANSI/API Standard 521, 2013 Table A.1).
where:
Q is the total heat absorption (input) to the wetted surface, API Committee plotted the total heat input versus the wetted area
expressed in W (Btu/h); from several pool re tests and an actual pool re (see API 521
C1 is a constant [ 43,200 in SI units (21,000 in USC units)]; Table A.1). The results, shown in Fig. 2, indicate the heat input
C2 is a constant [ 70,900 in SI units (34,500 in USC units)]. correlates with a 0.82 exponent on the wetted area (Aws). It should
F is an environment factor for reproong (F 1 for no be noted that, per convention, the heat input across only the wetted
reproong); surfaces of vessels containing a liquid that can boil is considered
Aws is the total wetted surface, expressed in square meters when designing pressure relief systems for the re scenario using
(square feet). the empirical method. The effect of heating unwetted surfaces and
Note 1: the SI equation constants include a conversion factor for gas-lled vessels is discussed later in this article.
(Aws)0.82. The effect of drainage was determined from Hottels pool re
Note 2: Inadequate drainage implies the burning liquid can test data (ANSI/API Standard 521, 2013, Table A.1, Test 1 and Test 2).
engulf the vessel resulting in increased heat input as compared Test 1 actually consists of the average of 31 tests without either
with a non-engulng type pool re where the burning liquid drains drainage or reghting. Test 2 consists of the average of 8 tests with
away from the vessel. drainage and 5 tests with both drainage and reghting. The ratio
Note 3: Fireghting reduces the re heat input to a vessel by for the pool re heat input from Test 2 to Test 1 (no drainage nor
water spray cooling of vessel surfaces. reghting) is 17,400/30,500 0.6. Hence, the maximum pool re
The origin of the empirical method can be traced back to the heat input to vessels with adequate drainage and reghting is
1950s, when the API Pressure Relief Systems (PRS) technical 34,500*0.6 21,000 BTU/h ft2 (see constant C1 in Equation (1)).
committee analyzed the available pool re test data and developed Hottels test data from Personal correspondence (1950), shown in
empirical equations to determine the pool re heat input to a vessel Table 1, suggest that drainage alone (no reghting) has a com-
(Heller, 1983). This heat input could then be used to calculate the parable reduction in heat input.
re relief load by dividing by the heat of vaporization. These The remaining parameter of the empirical method, designated
empirical equations include the: F, is the environment factor which credits for adequate re-
proong (See ANSI/API Standard 521, 2013). Fireproong that
 Maximum re heat input (i.e., maximum heat ux absorbed by meets the requirements of API 521 reduces the pool re heat input
the vessel and its contents) to a vessel thereby reducing the relief requirements. Fireproong
 Effect of wetted surface area of the vessel or equipment (i.e., area will also reduce the vessel wall heatup rate thereby increasing the
of the equipment in contact with liquid or below liquid level) on
re heat input
 Effect of burning liquid drainage (i.e., whether the pool re
engulfs the vessel) on re heat input

The maximum re heat input into a vessel is sometimes


confused with the ame surface heat ux (i.e., pool re heat duty
divided by ame surface area) and the incident heat ux at a vessel
exposed to the re. The incident heat ux excludes reduction in
heat ux due to the absorptivity of the vessel and re-radiation from
the vessel. Based on plotting hydrocarbon pool re test data (see
Fig. 1), the maximum heat input into the vessel was determined by
the API committee to be 34,500 BTU/h ft2 of wetted surface (see
constant C2 in Equation (2)). This maximum heat input would occur
when the wetted surfaces of the vessel are completely and
continuously exposed to ame.
Outdoor pool res are easily inuenced by even relatively calm
wind conditions; wind causes ames to move around, thereby only Fig. 2. API 521 Table A.1 re tests e Wetted area versus re heat input.
intermittently exposing surfaces of larger vessels to the highest Data sources: all tests shown in ANSI/API Standard 521 (2013) Table A.1 as well as the
incident heat ux. To determine the effect of vessel wetted area, the actual plant re involving a 380 butane sphere.
E. Zamejc / Journal of Loss Prevention in the Process Industries 27 (2014) 21e31 23

Table 1 Pressure relief systems designed for re exposure require a total


Hottel pool re test data showing effect of drainage with and without reghting heat input into the relevant surfaces (e.g., wetted areas). Because
(Personal correspondence, 1950).
the re heat intensity will vary with time and location, the total
Run Drainage Fireghting Average heat input, BTU/h ft2a heat input should be determined using a surface average heat ux
10 Yes None 22,862 obtained by averaging the re heat intensity across the entire ame
11 Yes None 30,081 volume. In contrast, when designing depressuring systems, the
12 Yes None 9819 peak re heat intensity (designated as the local peat heat ux) is
13 Yes None 19,517
important because localized overheating in a small area can result
14 Yes None 21,707
15 Yes None 26,953 in equipment failure due to overheating.
17 Yes None 2166 It should be noted that API 521 5th and prior editions provide
19 Yes None 16,436 design guidance for pressure relief and depressuring systems only
18 Yes Chemical foam for 30 s 6594
for open pool res. Also, the 34,500 Btu/h ft2 maximum re heat
then mechanical foam
22 Yes Chemical foam for 30 s 9337
input used in the API empirical method includes the vessel ab-
then mechanical foam sorptivity, which can reduce the incident heat ux by 20e70%.
23 Yes Chemical foam for 30 s 14,535 It is important to note that in the case of jet re exposure of
then mechanical foam vessels, a pressure relief device will usually not provide signicant
24 Yes Chemical foam for 30 s 25,028
protection/mitigation benet because the vessel walls can be
then mechanical foam
25 Yes Chemical foam for 30 s 20,672 heated by an impinging jet re to a temperature where the material
then mechanical foam loses strength resulting in vessel failure. In these cases, a properly
Average for all tests 17,362 designed depressuring system is one potential mitigation alterna-
Average for tests with 18,693 tive that can be used.
drainage, but no reghting
Average for tests with both 15,233
drainage and reghting
3. Analytical method to evaluate res
a
Heat input into wetted surface area.
Several authors recommended using an analytical method to
model the re scenario to overcome limitations with the empirical
time to failure of gas-lled vessels. The 6th edition has an expanded method and to provide more exibility in modeling (e.g. Energy
expression to account for multi-layer reproong. It is important to Institute, 2003; Roberts, Medonos, & Shirvill, 2000; Salater;
note that the API 521 criteria for determining adequacy of re- Salater, Overa, & Kjensjord, 2002; SCANDPOWER, 2004; Shirvill).
proong correspond to typical pool re exposure, but not jet res The analytical method is based on theoretical heat transfer equa-
because of the erosive power of the momentum jet. API 521 does tions that include radiative and convective heat transfer terms. In
not allow credit for re protection systems because of reliability contrast, the API 521 empirical method determines the re heat
aspects. input based on a correlation derived from pool re test data where
all of the heat transfer terms and effects are lumped together as
2.2. Limitations of the API empirical method to evaluate res shown in Equations (1) and (2). Because the analytical methods are
becoming more widely used, the API 521 committee agreed to
There are two types of res relevant to pressure relief and incorporate the analytical method as an alternative to the empirical
depressuring system design e pool re and jet re. A pool re is method and to include guidance on its application. The empirical
dened as a burning pool of liquid. A jet re is a re created method will still be recommended as the preferred method to
when a leak from a pressurized system ignites and forms a burning evaluate most re scenarios involving pressure relief system design
jet. A pool re can be classied as an open pool re, a conned pool with the analytical method preferred for special cases and res
re, or somewhere in between. A conned pool re is dened as a outside the scope of the empirical method.
re inside a building or a compact process module where the walls The analytical method to evaluate res is a basic heat transfer
and/or surrounding equipment can reradiate and preheat the equation as shown in Equation (3). The method determines the re
combustion air causing higher heat uxes than an unconned (i.e., heat input to a vessel and conservatively ignores internal heat
open) re. Generally only pool res are considered when transfer limitations. It can be applied to all types of res including
designing pressure relief systems while both pool res and jet res open pool res, conned pool res, and jet res. The proposed
are often considered when designing depressuring systems. typical ranges in the parameters are given in Table 2 for the surface
Typical ranges for peak re heat intensity (i.e., incident heat average heat ux for pool res. Recommended values, which
ux) are (Energy Institute, 2003): should be used where data or other resources are unavailable, are
given in ANSI/API Standard 521, 2013. ANSI/API Standard 521, 2013
. Open pool re e 16,000e48,000 Btu/h ft2 (50e150 kW/m2) also provides typical values of parameters for the local peak heat
. Conned pool re e 32,000e79,000 Btu/h ft2 (100e250 kW/m2) ux of pool res and for the local peak and surface average heat
. Jet re e 32,000e127,000 Btu/h ft2 (100e400 kW/m2) uxes for jet res.
 
These peak re intensities generally correspond to locations qabsorbed s$ asurface $fire $Tfire
4 4
 surface $Tsurface
within the re where the stoichiometric fuel-to-air ratio is equal to  
one. Because of the effects of ventilation (e.g., wind effects and h$ Tgas  Tsurface (3)
connement), fuel type, fuel-air stoichiometry and other factors, where:
the peak re heat intensity is generally only observed in localized qabsorbed is the absorbed heat ux from the re, expressed in Btu/
parts of the ame volume. Most of the ame volume has signi- h ft2 (W/m2);
cantly lower re intensities than the peak. Note that API 521 does s is the StefaneBoltzmann constant 0.1713  108 Btu/
not provide specic guidance on completely conned res because h ft2  R4 (5.67  108 W/m2 K4);
of their complexity (e.g., sensitivity to ventilation effects) which asurface is the equipment absorptivity, dimensionless;
generally requires case-by-case evaluations. re is the re emissivity, dimensionless;
24 E. Zamejc / Journal of Loss Prevention in the Process Industries 27 (2014) 21e31

Table 2 gallon (125 m3) LPG tank car was exposed to a semi-enclosed pool
Typical range in analytical method (Equation (3)) parameters for an open pool re re where the pool re and tank car were located in a pit, with
surface average heat ux.
embankments on all sides exceeding the tank car height (no roof).
Parameter Description Pool re surface average Fire and wall temperatures versus time at the top of the front and
heat ux parameter range rear walls of the tank car are shown in Fig. 3a and b, respectively.
re Hydrocarbon ame emissivity 0.6e1.0 The analytical method (Equation (3)) was used by the author in
surface Equipment emissivity 0.3e0.8 an attempt to reproduce the wall temperature versus time. The
asurface Equipment absorptivity 0.3e0.8
reported values in Anderson et al., 1974 were used for the param-
h Convective heat transfer 1.76e5.28 Btu/h ft2  R
coefcient between equipment (10e30 W/m2 K) eters where given (e.g., re temperature, re emissivity, initial
and surrounding air temperature). There is insufcient technical basis to theoretically
Tgas Temperature of combustion 1392e2112  R (932e1652  F) predict the value of the unknown parameters that would apply to
gases owing over the surface 773e1173 K (500e900  C)
this specic re due to the re variability. Hence, values for the
Tre Fire temperature 1572e2292  R (1112e1832  F)
873e1273 K (600e1000  C)
unspecied parameters were adjusted by trial-and-error until the
Tsurface Equipment temperature Increases as surface heats up calculated timeetemperature prole approximated the wall tem-
s StefaneBoltzmann constant 0.1713  108 Btu/h ft2  R4 perature data from the re test shown in Fig. 3b. The values of
(5.67  108 W/m2$K4) parameters selected to model tank car rear wall temperature versus
qre Fire heat ux e a wider range 9510e31,700 Btu/h ft2
time at several locations were:
is possible (30e100 kW/m2)
qabsorbed Absorbed heat ux at start 7925e23,775 Btu/h ft2
of the re (25e75 kW/m2) re 0.62 (determined by BRL)
surface 0.5
surface is the equipment emissivity, dimensionless;
asurface 0.5
h 1.76 Btu/h ft2  R (10 W/m2 K)
Tre is the re temperature, expressed in  R (K);
Tgas 2112  R (1652  F) 1173 K (900  C)
Tsurface is the equipment temperature, expressed in  R (K);
Tre 2112  R (1652  F) 1173 K (900  C) e see Fig. 3a
Tgas is the temperature of air/re in contact with the equipment
Tsurface Initially @ 529  R (69  F) 294 K (21  C)
surface, expressed in  R (K);
h is the convection heat transfer coefcient of air/re in contact
the equipment, Btu/h ft2  R (W/m2 K);
s$asurface$re$T4re is the radiative heat ux to the equipment;
s$surface$T4surface is the re-radiation from the equipment;
h$(Tgas  Tsurface) is the convection heat transfer between the
combustion gases and the equipments surface.
h is the convection heat transfer coefcient of air/re in contact
the equipment, Btu/h ft2  R (W/m2 K);
s$asurface$re$T4re is the radiative heat ux to the equipment;
s$surface$T4surface is the re-radiation from the equipment;
h$(Tgas  Tsurface) is the convection heat transfer between the
combustion gases and the equipments surface.
Note that published heat transfer coefcients and emissivities
are often empirically determined but the test conditions may not
accurately represent the conditions associated with a particular
re. Caution must be taken when specifying the parameters
because a wide range in re heat inputs can result. When applying
the analytical method to sizing pressure relief devices, the total
heat input into the vessel shall use the wetted area to the 1.0
exponent, not the 0.82 exponent used in the API empirical method
as shown in Equations (1) and (2). The 0.82 exponent is empirically
derived from re test data as shown in Fig. 2 and has no theoretical
basis. It is important to note that API 521 provides methods to
determine the re heat input to equipment while it is up to the user
to determine the vessel area exposed to a re based on the type,
size, conguration and location of their postulated re. Conse-
quence modeling is outside the scope of API 521.
Application of the analytical model to pool res is discussed in
Section 4.0 Application of the analytical method to modeling jet
res is given in Salater and Overa, (2004).

4. Application of the analytical method to model recent pool


re tests

4.1. Use of the analytical method to model the Ballistics Research


Laboratory (BRL) pool re test wall temperature versus time
Fig. 3. (a). Ballistics Research Laboratory pool re test data illustrating re tempera-
ture versus time at the top of the front and rear walls of a rail tank car. (b). Ballistics
In 1974, a pool re test was performed by the Ballistics Research Research Laboratory pool re test data illustrating rail tank car wall temperature
Laboratory (Anderson, Townsend, Zook, & Cowgill, 1974). A 33,000 versus time at the top of the front and rear walls.
E. Zamejc / Journal of Loss Prevention in the Process Industries 27 (2014) 21e31 25

qre Maximum of 16,700 Btu/h ft2 (52.7 kW/m2) (calculated)


qabsorbed Maximum of 9560 Btu/h ft2 (30.2 kW/m2) (calculated)

qre is calculated by setting surface 0, asurface 1. The maximum


absorbed heat ux is predicted by the analytical model to occur at
the start of re when the equipment is at ambient temperature. A
similar set of values was used by the author to predict the front wall
temperature versus time with the exception that the re and gas
temperature was set at 1472  F (800  C). Other combinations of
values in the analytical model can be used that may provide an
equal or better t to the test data. Note that, because the vessel was
engulfed in the pool re, the gas temperature should be set equal to
the re temperature. The gas temperature will be lower than the
re temperature for non-engulng pool res. Also, the re and gas
temperatures were assumed to be constant throughout the pool Fig. 5. BAM pool re test setup involving propane rail tank car (Balke et al., 1999;
Ludwig & Heller, 1999).
re.
A comparison of the analytical model with the wall tempera-
tures recorded during the test is shown in Fig. 4. The analytical  Embankment dimensions: 60  50  6 m (197  164  20 ft)
method provides a reasonable approximation to the observed rear  Tank car capacity 12,000 gallons (45.36 m3) TeT
wall temperature versus time. The leveling off of the front wall length 5.95 m; (19.5 ft) Diameter 2.9 m (9.5 ft)
temperature at about 800  F (425  C) as observed in the test (see  Tank car test pressure 28 bar (406 psia)
Fig. 3b) cannot be approximated with a single set of parameters,  Tank car contained liquid propane
indicating that one or more parameters changed during the course  Fuel oil pool re in troughs under the tank car and Castor
of the re. Once the temperature versus time prole is approxi- container
mated by the analytical method, then the resultant vessel heat  Castor container is used to store and transport radioactive ma-
input (i.e., qabsorbed) can be determined. The analytical method will terials and was tested along with the tank car.
indicate the maximum heat input is at the start of the pool re
where the vessel wall temperature is the lowest unless parameters in Figs. 5 and 6. Although the tank car was semi-conned by em-
change during the course of the re or there was an instrument bankments on 3 sides, a light to calm northerly wind was still able
fault. to signicantly affect the pool re exposure of the tank car as
shown in Fig. 7. The tank car maximum pressure reached 25 bar
4.2. Use of the analytical method to model the Federal Institute for (362 psig) about 15 min after the start of the pool re at which time
Materials Research and Testing (BAM) pool re test (Balke, Heller, the tank car ruptured, resulting in a boiling liquid expanding vapor
Konersmann, & Ludwig, 1999; Ludwig & Heller, 1999) wall explosion (BLEVE). The BLEVE aftermath is shown in Fig. 8. Pool re
temperature versus time ame/gas and tank car wall temperatures versus time at the various
locations around the tank car, as shown in Fig. 9, are illustrated in
In 1999, a full scale pool re test was performed by the Federal Figs. 10 and 11, respectively. It is important to note that failure
Institute for Materials Research and Testing (BAM or Bundesanstalt occurred before the pressure reached the pressure relief device
fr Materialforschung und -prfung) in Germany (Balke et al., 1999; opening pressure. A discussion of this failure as it relates to
Ludwig & Heller, 1999). The test evaluated and compared re depressuring system design is given in Section 6.3.
exposure effects on a rail tank car containing propane and a Castor Because of the wide pool re and wall temperature ranges
car used to transport radioactive material. The test setup is shown shown in Figs. 10 and 11, a single set of values for the parameters in
the analytical method would not predict all variations. Only tem-
perature data was given in the report. As in the BRL comparisons,

Fig. 4. Comparison of rail tank car wall temperature versus time between the Fig. 6. BAM pool re test setup involving propane rail tank car (Balke et al., 1999;
analytical model and Ballistics Research Laboratory pool re test data. Ludwig & Heller, 1999).
26 E. Zamejc / Journal of Loss Prevention in the Process Industries 27 (2014) 21e31

test data. The rail tank car was lled with about 2650 gallons
(10 m3) of 95% liquid propane, resulting in an initial wetted surface
area of about 249 ft2 (23.16 m2). The pool re heat input deter-
mined by the author for the empirical method without adequate
drainage (see Equation (2)), and the analytical method for several
locations around the tank car are given in Table 4. There are two
locations in the rear of the tank car where the analytical method
indicated higher heat inputs than the empirical method. However,
when averaged across the entire tank car, as one should do if sizing
a pressure relief device, the empirical method resulted in about 30%
more heat input than the analytical method.

5.2. Comparison with pool re heat input based on BAM re test


liquid sensible heating

Test data on the sensible heating of the propane liquid was


obtained during the BAM test. This data can be used as an inde-
Fig. 7. BAM pool re test (Balke et al., 1999; Ludwig & Heller, 1999). Note: taken near
pendent means to determine pool re heat input during the BAM
end of test (calm wind speed).
test. Note that the pressure did not reach the pressure relief device
unspecied parameters had to be determined by the author by opening pressure prior to rail tank car failure during the BAM test.
trial-and-error whereby values were selected so the calculated The test data indicated an average temperature rise of 8.06  F/min
time-versus-temperature prole matched the re test data as close (4.48  C/min). Hence, the calculated total heat input due to sensible
as possible. Table 3 illustrates the values of parameters selected to heating of the liquid is about 3.805  106 BTU/h (1115 kW). For
model tank car wall temperature versus time at two locations. comparison, the empirical method (assuming inadequate drainage)
Other combinations of values in the analytical model can be used predicted a total heat input of 3.18  106 BTU/h (933 kW) per
that may provide an equal or better t to the test data. It should be Table 4. This is roughly the same as the analytical approach using
noted that a transient approach to the analytical method, where the only the averaged tank car rear temperature data. A possible reason
re and gas temperatures were varied with time based on the test for these differences is discussed below.
data shown in Fig. 10, was evaluated; however, did not appear to Liquid swelling as the liquid heats up would increase wetted
signicantly improve the t with the test data. surface area; however, the temperature did not increase enough
A comparison of the analytical model with the wall tempera- during the test for it to explain the difference between the test data
tures recorded during the test, shown in Fig. 12, indicates the and the empirical and analytical methods. A likely explanation is
analytical method can provide reasonable approximations to wall that the embankment on three sides of the tank car heated up
temperatures versus time if parameter values are empirically during the re and caused higher heat uxes due to re-radiation,
determined by selecting those that would provide a temperature- preheating of combustion air, and enhanced heat transfer. Indeed,
time prole similar to the test data. the re should be classied as semi-conned because the height of
the embankment walls exceeded the height of the tank car. In such
cases, the API 521 empirical method (Equations (1) and (2)) does
5. Comparison of the pool re heat inputs between the
not directly apply. However, the equations can be modied by using
empirical method, the analytical method and pool re test
a wetted surface area (Aw) exponent of 1.0 instead of 0.82. This
data
would be appropriate in scenarios where the pool re ames
directly and continuously contact all of the wetted surfaces.
5.1. Comparison with pool re heat input based on BAM time-
Applying this to the BAM test rail tank car assuming 50% of the rail
versus-temperature test data
tank car is partially conned due to the embankments on three
sides results in the Equation (4):
The pool re heat input determined by the empirical method
and the analytical method can be compared using the BAM pool re

Fig. 9. BAM pool re test e Temperatures measurement locations (Balke et al., 1999;
Fig. 8. BAM pool re test e BLEVE aftermath (Balke et al., 1999; Ludwig & Heller, 1999). Ludwig & Heller, 1999).
E. Zamejc / Journal of Loss Prevention in the Process Industries 27 (2014) 21e31 27

situations (e.g., enclosed buildings or structures with a roof) which


would require special modeling. Note that the empirical method is
based on full-scale test data, not theory. Hence, this adjustment of
the wetted exponent for partially conned vessels based on test
data is consistent with the origin of the empirical method.

5.3. Comparison with pool re heat input based on the BRL test

The BRL test obtained data on the relieving rate versus time, which
was compared with that obtained with the empirical and analytical
methods. A transient approach was used by the author in these
methods whereby the relief rate was varied with time to correspond
to the decrease in wetted area as uid is relieved. A comparison of the
actual relief rate and that predicted by the empirical and analytical
methods is given in Fig.13. Both the empirical and analytical methods
predicted a decrease in relief rate with time because the wetted area
is decreasing as uid is relieved. However, the test indicated the relief
rate actually increased with time.
Fig. 10. BAM pool re test e Fire temperatures versus time (Balke et al., 1999; Ludwig One explanation that can increase the relief rate versus time is
& Heller, 1999). Note 1: Time 0 is when gasoline starter uid in a small plastic that there was an increase in heat ux with time due to heating of
container was ignited. The main pool re started about 100 s later when the plastic
the surroundings. Adjustments to the analytical method were made
container failed and spilled burning gasoline into the fuel oil pool. Note 2: The un-
marked temperature curves were primarily in the front of the tank car (upwind to account for enhanced heat transfer due to heat-up of the sur-
location and without an adjacent embankment). rounding embankments during the test. Fig. 13 illustrates a modi-
ed analytical method where the convective heat transfer
coefcient and the vessel absorptivity were increased by 20% every
Q API modified empirical method 2.5 min with a limit of 1.0 for the absorptivity. These were empir-
h i
70; 900* Aw confined1:0 Aw open0:82 (4) ically determined by trial-and-error so predicted relief rate versus
time matched the re test data as close as possible. This should be
h i considered an example of adjustments that can be made to adjust
Q API 70; 900* 11:581:0 11:580:82 the model to t test data, but they may not represent actual con-
ditions nor be applicable to other res. Note that the re and gas
1:349  106 Watts 4:604  106 BTU=h temperature were not adjusted because the data showed the re
temperature to slightly decrease during the test.
This is a conservative estimate of the total heat input as compared
Instead of modifying the analytical method, the empirical
with the 3.805  106 BTU/h (1115 kW) determined from liquid
method can be adjusted to account for the apparent increased heat
sensible heating. Based on the test data, the analytical method
input with time by increasing the exponent on the wetted area
should use the rear averaged heat input predicted by the analytical
versus time. This effect can be illustrated by inserting the heat input
model (i.e., 3.75  106 BTU/h (1098 kW)) to obtain a reasonable
determined from the actual relief rate and the wetted area in the
approximation. Where validating data is unavailable, the highest
empirical method for inadequate drainage; the equation is then
heat input obtained from the analytical model should be used.
solved for the wetted area exponent versus time. The results,
The re relief load can be determined by dividing the re heat
shown in Fig. 14, indicate the wetted area exponent approaches 1.0
input by the heat of vaporization of the uid at relieving pressure.
toward the end of the test. This suggests that ame contact with the
These results indicate that the API empirical method can be
entire vessel surfaces increases with time. Using a wetted area
applied to some semi-conned congurations, where adjacent
exponent of 1.0 for the entire vessel (located in a pit with em-
embankments exceed the vessel height, by using a wetted area
bankments exceeding the vessel height on all sides), with the
exponent of 1.0 instead of 0.82 for the portion of the vessel adjacent
empirical method, would provide a conservative pressure relief
to the embankment. This would not apply to completely conned
system design.

5.4. Application to pressure relief device sizing

The analytical model can be used to estimate the time-versus-


temperature proles of unwetted walls of vessel exposed to pool
res. API 521 provides recommended values for the parameters in
this model that should provide a reasonable approximation for a
wide range of pool res. However, the comparisons with full-scale
re tests indicate the model parameters would need empirical
adjustment based on actual full-scale test data in order to more
closely predict a specic pool re behavior. Unfortunately, the
state-of-the art is insufcient to allow a validated theoretical basis
for adjustment that would apply to pool res in general. Hence, the
user is cautioned when applying the analytical model to pressure
relief device sizing for the re exposure scenario because improper
selection of the parameter values can result in underpredicting the
Fig. 11. BAM pool re test e Tank car wall temperatures versus time (Balke et al., 1999; re heat input to a vessel and consequent undersizing of the
Ludwig & Heller, 1999). pressure relief device. Where validating re test data is unavailable,
28 E. Zamejc / Journal of Loss Prevention in the Process Industries 27 (2014) 21e31

Table 3 that metals such as carbon steel lose signicant strength. Table 5
Analytical method parameters used to model tank car wall temperature versus time illustrates the effect of high temperatures on the tensile strength
at two locations of the BAM pool re test tank car.
of carbon steel and 304 stainless steel. The loss of strength due to
Parameter Rear center Front right pool or jet re exposure could exceed the safety factor used in the
Fire temperature,  F ( C) 1832 (1000) 482 (250) design of the vessels, thereby resulting in vessel rupture due to
Gas temperature,  F ( C) 1832 (1000) 482 (250) overheating, rather than overpressure.
Convective heat transfer 3.52 (20) 3.52 (20) The specic pressure vessel design code and material used will
2  2
Coefcient, BTU/h ft R (W/m K)
determine the appropriate safety factor to use in the vessel design.
Fire emissivity 0.6 0.6
Metal emissivity 0.5 0.4 For example, the current edition of ASME Section VIII, Division 1
Metal absorptivity 0.5 0.4 Pressure Vessel Design Code (ASME Section VIII) includes a safety
The following were calculated using the values above: factor (now termed design margin) of 3.5 between the tensile
Calculated initial incident heat ux, 34,570 (109.0) 2260 (7.1)
strength of the vessel and the allowable stress at room temperature
BTU/h ft2 (kW/m2)
Calculated maximum absorbed heat ux, 20,321 (64.1) 1721 (5.4)
for materials in which the tensile strength governs (e.g., carbon
BTU/h ft2 (kW/m2) steel). For carbon steel, the safety factor implies the design pressure
is a minimum of 3.5 times the burst pressure (assuming the weak
link in the vessel is the wall plate, there are no imperfections in the
the user should use the API 521 empirical method for sizing pres- wall, etc.). It should be noted that carbon steel vessels constructed
sure relief devices for the pool re scenario. to pre-1999 versions of the ASME Section VIII, Div. 1 code used a
The comparisons in the previous sections indicate the 50 year safety factor of 4.
old empirical method in API 521 given in Equations (1) and (2) would ASME Section VIII, Division 1, UG-27 provides equations that
provide a reasonable approximation of unconned pool re heat relate the allowable stress, vessel design pressure and wall thick-
input to a vessel but would need adjustment for cases where the ness. In the case of circumferential stress for a cylindrical shell,
vessel is partially conned. If the vessel is partially conned, the BRL Equation (5) applies in many cases:
full-scale test indicates a conservative size for a pressure relief de-
vice can be obtained using the heat input from Equation (2) but with P S*E*t=R 0:6*t (5)
a wetted area exponent of 1.0 instead of the 0.82 exponent valid for
unconned pool res as shown in Equation (4). where:
API 521 does not provide any recommendations regarding P must be < 0.385 *S*E.
completely conned pool res. In the case of completely conned
pool res, there is no full scale test data available so it cannot be P internal design pressure, psi (MPa)
determined if the API 521 empirical methods would be conservative. E joint efciency 1.0 for full X-ray.
Although the analytical method may provide the potential to model S maximum allowable stress value, psi (MPa)
completely conned pool res, selection of the values for the model t minimum thickness of the shell, inches (mm)
parameters would need to consider their variation with time as the R inside radius of the shell, inches (mm)
re heats the surroundings. This can only be done on a case-by-case
basis. Further, validation tests to support selection of appropriate For example, at room temperature, ASTM A515 Grade 70 carbon
values for the model parameters would be recommended. steel plate has a tensile strength of 70,000 psi (482.6 MPa) (ASME
Section II, 2007) therefore, the allowable stress will be
S 70,000/3.5 20,000 psi (137.9 MPa). If a vessel fabricated from
6. Application of the analytical method to depressuring
this material and designed to this allowable stress is heated to
system design
1200  F (650  C), the tensile strength will decrease to 20,000 psi
(137.9 MPa), as shown in Table 5. In other words, the material
6.1. Effect of overheating unwetted metal plates
strength is reduced to the equivalent of a zero safety factor. Vessel
rupture would be a certainty if the pressure then exceeded the
Unwetted metal surfaces are not cooled by boiling liquid inside
design pressure because the loads on the vessel would exceed the
the vessel. Hence, the metal temperature can get high enough such
tensile strength. Rupture would occur at even lower internal
pressures if there are other coincidental loadings on the vessel
(such as the weight of the vessel and attached equipment, tem-
perature gradients, static head, internals, etc.) or defects in the
vessel. In all these cases, a pressure relief valve would not provide
protection because it is typically designed to reseat (i.e., close)
when the pressure decreases to about 93% of its set pressure for
vapor trim valves. This would thereby maintain the vessel pressure
near its design pressure. Instead of a pressure relief valve, a
depressuring system can be used to provide vessel protection, or at
least mitigation of the effects of failure.

6.2. Depressuring criteria

In order to be effective, the depressuring system needs to


depressure at a high enough rate to compensate for the loss of
strength as the vessel heats up. The vessel heat up rate is dependent
on the type of re, materials of construction, and wall thickness.
Fig. 12. Comparison of rail tank car wall temperature versus time between the API Std. 521 Fig. 1 illustrates the heat-up of carbon steel plates of
analytical model and BAM test data. several thicknesses in an open pool re (ANSI/API Standard 521,
E. Zamejc / Journal of Loss Prevention in the Process Industries 27 (2014) 21e31 29

Table 4
Pool re heat inputs using the empirical and analytical method along with BAM re test data.

Location Max re heat ux, Max absorbed heat ux, Aw exponent Total heat input, BTU/h (kW) % Of API
BTU/ft2 h (kW/m2) BTU/ft2 h (kW/m2)

Analytical e Rear center 34,560 (109) 20,330 (64.1) 1 5.07  106 (1485) 159%
Analytical e Front right 2260 (7.1) 1720 (5.4) 1 0.43  106 (126) 14%
Analytical e Rear average (Note 1) 24,350 (76.79) 15,040 (47.41) 1 3.75  106 (1098) 118%
Analytical e Front average (Note 1) 6650 (20.97) 4620 (14.57) 1 1.15  106 (337) 36%
Analytical e Total average 15,500 (48.88) 9830 (30.99) 1 2.45  106 (718) 77%
Empirical method N/A 34,500 (70.9) (Note 2) 0.82 3.18  106 (933) 100%

Note 1: Average of left, center and right locations.


Note 2: The API maximum absorbed heat ux has units of BTU/h/[(ft2)0.82] or kW/[(m2)0.82].

2013). One curve (Plate 2) was obtained from open pool re test 6.3. Application of the analytical method to depressuring system
data while the others were extrapolated based on the test data. design
Combining these temperature-versus-time curves along with the
tensile strength data shown in Table 5 will allow determination of a The analytical method can be used to extend the curves in Fig. 15
minimum depressuring rate to keep the pressure below the tensile to other wall thicknesses. The analytical method along with the
strength of the vessel. An appropriate safety factor should be parameters determined in Section 6.2, for example, can be set up in
considered given the uncertainties. Results obtained by the author, a spreadsheet as a transient model in which the wall temperature
applying a 25% safety factor (i.e., Table 5 tensile strengths were change with time is calculated. At each time interval, the metal wall
multiplied by 0.75), are shown in Fig. 15. The depressuring prole mass can be conservatively assumed to absorb all of the heat input,
for a specic wall thickness needs to stay to the left of the specic thereby increasing the wall temperature. The effect of wall thick-
curve shown in Fig. 15. Failure will occur if the depressuring prole ness is accounted by the metal mass. This temperature-versus time
either intersects or is on the right side of the curve for the thickness prole is then combined with tensile strength data as in Section 6.2.
in question. As noted in the previous section, failure can occur at For example, the BAM pool re test data indicated failure of the
even lower pressures, depending upon the amount of additional rail tank car occurred at rear center wall (in unwetted zone) (Balke
loads on the vessel. et al., 1999; Ludwig & Heller, 1999). Test data further indicated the
An often used criteria for depressuring is to depressure to 50% of wall temperature ranged from 1020 to 1200  F (550e650  C), but it
the design pressure in 15 min. As shown in Fig. 15, this would be is possible that local temperatures got even higher because tem-
appropriate for open pool re exposure of vessels whose wall perature was recorded only at a few locations. Failure occurred
thickness is 1 inch or greater. A second criteria often used is to 15 min after the start of the pool re, or about 10 min after the re
depressure to 100 psig (6.90 barg) in 15 min. This is generally more temperature reached about 1832  F (1000  C). The rail tank car wall
conservative given the high pressures involved in most depres- thickness was 0.59 inches (14.9 mm) and the material of con-
suring applications. The more stringent criteria, (almost always the struction was assumed to be comparable to carbon steel. The failure
latter) would be preferred when protecting against jet re pressure of 362 psig (25 bar) was slightly lower than the test
exposure. pressure of 406 psig (28 bar). The Rear Center parameters were
used in the analytical model to predict the time-versus-
temperature prole. This was combined with the tensile strength
and stress rupture data by the author to obtain the depressuring
prole shown in Fig. 16. In order to minimize the potential for
rupture due to overheating, a depressuring system would need to
stay to the left of the curve shown in Fig. 16. Because the pressure at
failure was slightly lower than the test pressure, Fig. 16 predicts that

Fig. 13. Empirical and analytical method calculated relief rates versus BRL test data. Fig. 14. Empirical method wetted area exponent versus time using BRL test data.
30 E. Zamejc / Journal of Loss Prevention in the Process Industries 27 (2014) 21e31

Table 5
High temperature tensile strength of carbon steel and 18-8 stainless steel (Wharton,
1946).

Temperature Temperature 18-8 Stainless Carbon steel


 
F C steel (304, 304L) (SA-515, SA-516)

Tensile Tensile Tensile Tensile


strength strength strength strength
psi MPa psi MPa

900 482 45,500 313.7


1000 538 53,000 365.4 36,500 251.7
1100 593 48,500 334.4 27,200 187.5
1200 649 43,000 296.5 20,000 137.9
1300 704 35,000 241.3 13,500 93.1
1400 760 27,000 186.2 9025 62.2
1500 816 20,500 141.3
1600 871 17,650 121.7
Fig. 16. Depressuring prole to minimize failure potential of the rail tank car due to
overheating in the BAM pool re test.

failure would occur about 14 min after the start of the pool re,
which is a reasonable approximation as failure actually occurred
about 15 min after the main pool re started (see Figs. 10 and 11). committee to include an analytical method in the 6th edition as
Note the rst 2 min of the pool re test is not considered because an alternative to the existing empirical method. The analytical
the re was localized to a small igniter assembly that did not cause method provides more exibility than the empirical method but
any signicant increase in rail tank temperatures. has limitations (e.g., too many permutations are possible). API
521 provides recommended values that would apply to many
open pool res. However, the comparisons with full-scale pool
6.4. Effect of material of construction re tests indicate that caution needs to be taken when selecting
the values; otherwise, the re heat input can be underestimated
The material of construction can signicantly affect the resulting in an undersized pressure relief device. Where uncer-
depressuring requirements. The preceding sections discussed car- tain, the values selected in the analytical model should be vali-
bon steel vessels. As shown in Table 5, 304 stainless steel is superior dated with testing.
to carbon steel regarding high temperature effects on tensile More recent pool re test data indicates the 50 year old API
strength. A comparison of the depressuring proles to minimize 521 empirical method will provide a conservative estimate of pool
the potential for failure of a inch wall thickness carbon steel re heat input for most applications and is still the method of
vessel and a inch wall thickness stainless steel vessel is illustrated choice when designing pressure relief systems for an open pool re
in Fig. 17. The depressuring system pressure-versus time prole scenario. However, these recent tests indicate the empirical
would need to stay to the left of the applicable curve. These results method needs to be modied when a vessel or equipment is
indicate that the depressuring system for the stainless steel vessel partially conned by adjacent embankments or walls equal or
would require a signicantly lower depressuring rate than for the greater than the vessel height. In such cases, the wetted area
carbon steel vessel of comparable wall thickness. This method can exponent should be 1.0 instead of 0.82.
be extended to other materials provided tensile strength data at The analytical method is useful in determining time-versus-
high temperature is available. temperature proles for heating unwetted vessels of varying wall
thicknesses and materials of construction. These proles can be
7. Conclusions combined with tensile strength and stress-rupture data to specify a
depressuring systems pressure-versus-time prole to minimize
The increasingly widespread use of analytical methods to failure and/or mitigate the effects of failure due to overheating from
evaluate re exposure of equipment prompted the API Std. 521 a pool or jet re exposure.

Fig. 15. Reduction of carbon steel plate tensile strength versus time due to open pool Fig. 17. Depressuring proles to minimize failure potential of a 0.5 inch wall thickness
re exposure. carbon steel and stainless steel vessel due to overheating.
E. Zamejc / Journal of Loss Prevention in the Process Industries 27 (2014) 21e31 31

References National Fire Protection Association NFPA 30 (2008): Flammable and Combustible
Liquids Code.
Personal correspondence from H.C. Hottel to L.W.T. Cummings December 12, 1950.
Anderson, C., Townsend, W., Zook, J., & Cowgill, G. (September 1974). The effects of a
Roberts, T. A., Medonos, S., & Shirvill, L. C. (June 2000). Review of the response of
re environment on a rail tank car lled with LPG. FRA-OR&D Report Number 75e
pressurised process vessels and equipment to re attack. Offshore Technology
31, PB-241358.
report, OTO 2000-051.
ANSI/API Standard 2000. (November 2009). Venting atmospheric and low-pressure
Salater, P. Proposed changes to the next revision of API 521. 2006 Presentation to API
storage tanks (6th ed.).
Pressure Relief Systems Committee.
ANSI/API Standard 521. (2013). Pressure-relieving and depressuring systems (6th ed.).
Salater, P., & Overa, S. J. (March 2004). Pipes exposed to medium sized jet res e
API Standard 650. (2013). Welded tanks for oil storage (12th ed.).
Rupture conditions and models for predicting time to rupture. Paper presented at
ASME Section II, Part D, materials e Properties. (2007).
Fire and Blast Information Group (FABIG), London and Aberdeen, January 2004
ASME Section VIII, Division 1, Pressure Vessel Code, 2007 with 2008a Addenda.
and Houston.
Balke, C., Heller, W., Konersmann, R., & Ludwig, J. (September 13, 1999). Study of the
Salater, P., Overa, S. J., & Kjensjord, E. (September 2002). Size depressurization and
failure limits of a railway tank car lled with liqueed petroleum gas subjected to
relief devices for pressurized segments exposed to re. Chemical Engineering
an open pool re test. BAM Final Report.
Progress, 38.
Energy Institute. (March 2003). Guidelines for the design and protection of pressure
SCANDPOWER. (March 31, 2004). Guidelines for the protection of pressurised systems
systems to withstand severe res, ISBN 0 85293 279 0.
exposed to re. Report no. 27.207.291/R1-Version 2.
Heller, F. J. (1983). Safety relief valve sizing: API versus CGA requirements plus a
Shirvill, L. C. Heat Fluxes in Severe Fires. 2002 Presentation to API Pressure Relief
new concept for tank cars. API Rening Proceedings, 82, 123e140.
Systems Committee.
Ludwig, J., & Heller, W. (1999). Fire test with a propane tank car. BAM Test Report
Wharton, H. R. (1946). Digest of steels for high-temperature service. Timken Steel.
III.2/9907.

Vous aimerez peut-être aussi