Journal of Loss Prevention in the Process Industries 50 (2017) 154e164

Contents lists available at ScienceDirect

Journal of Loss Prevention in the Process Industries

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

Application and validation of the API analytical fire method in

pressure-relieving and depressuring systems
Souvik Biswas*, Benjamin J. Fischer
Aspen Technology, Inc., Bedford, MA 01730, USA

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

Article history: Accounting for fire scenarios is critical when analyzing pressure-relieving and depressuring systems. This
Received 10 August 2017 is particularly true in systems such as oil refineries, petrochemical facilities, gas plants, and oil and gas
Received in revised form production facilities where the flowing fluids are highly flammable. Because operating safely is of
14 September 2017
paramount importance in these industries, standards and recommended practices have been developed
Accepted 17 September 2017
by trade associations such as the American Petroleum Institute Standard 521 (API 521) to aid in the
Available online 21 September 2017
analysis of the system. In the case of fire scenarios, the traditional recommendation of API 521 has been
an empirical model based on the wetted area of the vessel. However, in the most recent 6th edition
Keywords:
Blowdown
released in 2014, the standard added an analytical equation based on the Stefan-Boltzmann law that can
Pressure vessel be used for modeling the dynamic response of pressurized vessels in fire scenarios. Using BLOWDOWN™
Hydrocarbons Technology in Aspen HYSYS, this analytical fire equation was validated with available experimental data
Fire and compared with the traditional wetted area model. This also included evaluating the effects of the
Simulation parameters in the analytical equation such as surface emissivity of the equipment wall and external heat
transfer coefficient, which have significant uncertainties. As plant fires can impact a wide area, this study
further investigated using the analytical equation to model fire on a detailed geometry of a pressurized
vessel with associated piping.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction addition, the fluid will follow a thermodynamic trajectory through

the phase space that will often include phase changes and be close
Pressurized vessels exposed to fire in a hydrocarbon processing to or in the critical region. A simulation of this process must
plant are an important area of investigation in the field of process therefore predict the time-dependent pressures, temperatures and
safety. When a fire occurs in a plant, it is critical to evacuate the multi-phase compositions within a vessel, temperatures of the
contents of any equipment in the vicinity in order to avoid esca- wall, and rates of efflux. Haque et al. (1992a) built a rigorous model
lation due to overpressure and potential catastrophic failure. A safe to study these issues. Their computer program BLOWDOWN
depressuring system should therefore be designed by including a accurately captures both the thermodynamic behavior of the multi-
careful analysis of fire scenarios. Often, this analysis can be per- component fluid inside the pressurized vessel and the heat transfer
formed using dynamic models. To be successful, the model must between the external thermal environment, the vessel wall and the
accurately represent the physical processes that occur inside the internal fluid. The results were experimentally validated under cold
vessel as well as the physical nature of the fire. conditions by tests conducted at Imperial College, London and the
During a depressuring event, a multiphase mixture of gas, liquid, British Gas facilities (Haque et al., 1992b). This technology is now
and aqueous phases will flow through the vessel, valve, and any part of the general-purpose process simulation tool Aspen HYSYS.
associated piping. As the fluid flows, there will be heat transfer Another important consideration is deciding how to model the
between the fluid and vessel wall as well as between the different behavior of the fire. Specifically, a model needs to be chosen for
fluid phases. There will also be an accompanying mass transfer how the additional heat flux from the external fire is applied to the
between the phases as liquid evaporates and vapor condenses. In equipment. For example, the heat flux of a pool fire can be signif-
icantly different from that of a jet fire. Traditionally, to design a
pressure relieving system to mitigate the risk of a fire hazard,
* Corresponding author. standard industry practice for modelling the fire is to use the
E-mail address: Souvik.Biswas@aspentech.com (S. Biswas).

https://doi.org/10.1016/j.jlp.2017.09.008
0950-4230/© 2017 Elsevier Ltd. All rights reserved.
 S. Biswas, B.J. Fischer / Journal of Loss Prevention in the Process Industries 50 (2017) 154e164 155

American Petroleum Institute Standard 521 (API 521) pool fire embankments or walls. Melhem and Gaydos (2015) found that this
equation. This is an empirical equation for the additional heat flux equation needs to be corrected by a fuel factor, which depends on
that should be applied to the liquid phase of the system. Conse- the type of fuel as the emissive power of fire caused by lighter
quently, this method is not recommended for determining the hydrocarbons is much greater than that by heavier hydrocarbons.
temperature response in the walls. To analyze the wall tempera- As mentioned earlier, Salater et al. (2002) and the Scandpower
tures and potential rupture of a cylindrical vessel that is fully- Risk Management AS (2004) provided a method on how to
engulfed by a fire, Mahgerefteh et al. (2002) simulated blowdown design such systems using the Stefan-Boltzmann radiation equa-
by modeling the fire as a constant heat flux to the outer wall of a tion to model a fire. This method has since been incorporated into
vessel. As a constant heat flux model may not capture all of the ANSI/API Standard 521 6th Edition (2014) as an alternative fire
physics of the fire and could be overly conservative, Salater et al. equation for designing for risk mitigation of a fire hazard in a
(2002) and the Scandpower Risk Management AS (2004) recom- process plant. Using this method, an additional heat flux to the
mended modeling the fire with the Stefan-Boltzmann equation for outer wall of the vessel or pipe is determined rather than an
thermal radiation along with the convective heat transfer from the additional heat flux to the liquid phase. The heat flux going into the
hot gases to the vessel. These recommendations also introduced the fluid inside the vessel is then calculated by solving the heat transfer
approach of using a global average heat flux with a local peak heat equations between the wall and the fluid inside the vessel. This
flux to account for the spatial and temporal variability associated heat flux is given by:
with any real fire.  
As previously mentioned, there is some experimental data in qabsorbed ¼ s as εf Tf4  εs Ts4 þ hðTa  Ts Þ (2)
literature on scheduled depressurization cases (Haque et al., 1992b)
that has been used to validate depressuring models under cold where
conditions. Validating models for emergency depressurization
scenarios where a vessel is under fire is challenging due to a lack of qabsorbed ¼ absorbed heat flux by the vessel wall [W/m2]
experimental data. However, there is data in the literature (Moodie s ¼ Stefan-Boltzmann constant [5.67  108 W/m2-K4]
et al., 1985, 1988) that examines the fire engulfment of LPG tanks. Tf ¼ fire temperature [K]
This data is used in the present study to validate the API analytical Ts ¼ surface temperature [K]
fire equation using the BLOWDOWN Technology in Aspen HYSYS. Ta ¼ ambient temperature [K]
Subsequently, several case studies of emergency depressurization εf ¼ fire emissivity
of vessels under fire attack are presented to compare the traditional εs ¼ surface emissivity of the vessel wall
API equation with the API analytical fire equation as well as to as ¼ surface absorptivity of the vessel wall
investigate the effects of the adjustable parameters in the API h ¼ heat transfer coefficient between surface and environment
analytical fire equation. Finally, a more detailed geometry that in- [W/m2-K]
cludes process piping is investigated in order to examine the effects
of local peak heat fluxes. The main advantage Eq. (2) is that it is based on first principles;
therefore, it has a much wider range of applicability and is not
limited to pool fires or wetted vessels. In addition, given accurate
2. Fire modelling using API 521 inputs about the fire, it can be used to accurately predict wall
temperatures of the vessel. Finally, it can be used to study the
To size an orifice associated with a blowdown valve used for integrity of vessels and piping.
emergency depressuring in a hydrocarbon processing plant, an One potential issue of using Eq. (2) to model fire scenarios is that
accurate model of how much heat will be absorbed from the fire is fires are often not uniform, which implies that there is not a uni-
required. Currently, the standard practice among safety engineers is form heat load from the fire to the vessel. To solve this issue, a
to follow the American Petroleum Institute Standard 521 for pres- recommended approach is to use a global average heat flux with a
sure relieving and depressurizing systems. In the traditional API local peak heat flux (Scandpower Risk Management AS, 2004;
empirical equation, the total heat absorbed is a function of the ANSI/API Standard 521 6th Edition, 2014). The global average heat
wetted area given by: flux is the spatial and temporal average of the heat flux over the
surface of the vessel. When multiplied by the vessel area, this is a
Q ¼ C1 FACw2 (1)
good indicator of the total heat absorbed by the whole vessel over a
 period of time. Consequently, it is suitable for analyzing the pres-
43:2 ðwith drainage and firefightingÞ sure profile inside the vessel. The local peak heat flux is the
C1 ¼ ;C2 ¼ 0:82
70:8 ðwithout drainage and firefightingÞ maximum heat flux at a point in the vessel surface, which can be
where used to estimate the conservative transient temperature response
of the vessel wall.
Q ¼ heat absorbed by the liquid inside the vessel [kW] Another difficulty in applying Eq. (2) to a fire model is that the
Aw ¼ wetted area [m2] parameters are not commonly known during the design and sizing
F ¼ environmental factor stage. To remedy this, API 521 provides ranges and default values
for these parameters. However, as noted by Zamejc (2014), it is still
As explained earlier, this heat is absorbed by the liquid inside of possible to apply this equation incorrectly due to the many allow-
the vessel and is not applied to the vessel wall. Therefore, it should able permutations of these parameters.
not be used to determine temperatures in the vessel wall nor is it
applicable for unwetted vessels.
Due to the empirical nature of this model, it may not be appli- 3. Validation of fire engulfment of LPG tanks with
cable under all conditions. Roberts et al. (2004) found this approach experimental results
to be non-conservative in cases of severe fires in offshore and in
some onshore situations. Zamejc (2014) recommended a value of Presently there is a lack of published experimental data for a
1.0 for the factor C2 when the vessel is partially confined by vessel undergoing emergency depressurization under fire. Most of
156 S. Biswas, B.J. Fischer / Journal of Loss Prevention in the Process Industries 50 (2017) 154e164

the available experimental results of vessels exposed to fire are

studies involving the fire engulfment of LPG tanks. The main pur-
pose of these studies was to either evaluate the effectiveness of the
pressure relief device protecting them or to estimate the time to
failure of the vessel. Anderson et al. (1974) performed a detailed
study of pool fire tests of a rail tank car filled with LPG. However,
the liquid levels inside the tank during the tests were not reported.
More recently, a full-scale pool fire test was conducted at the
Federal Institute for Materials Research and Testing (BAM) in Ger-
many (Balke et al., 1999; Ludwig and Heller, 1999). The main pur-
poses of this test were to evaluate and compare fire exposure
effects on a rail tank car containing propane and to measure the
time required to reach the boiling liquid expanding vapor explosion
(BLEVE) from fire ignition. In this time-to-failure test, full fire
engulfment was not achieved due to prevailing wind conditions. Fig. 2. Using the API 521 analytical fire equation, the heat flux is applied uniformly to
The previously mentioned experiments all lack some data that the outside vessel wall, which leads to varying heat fluxes applied to each phase in the
prevents a meaningful comparison with a model. However, tests vessel as a function of time.

carried out by the UK health and safety executive (Moodie et al.,

1985, 1988) provide a complete set of data for the full fire engulf-
decreases with time. Conversely, the heat flux to the wetted region
ment of LPG tanks of several sizes (0.25 ton, 1 ton and 5 ton). The
will remain relatively constant.
main objectives were to characterize the fire, determine the pres-
The use of Eq. (2) to model fire scenarios can be validated by
sure increase inside of the vessel, and evaluate the operation of the
comparing BLOWDOWN simulations with the fire engulfment tests
pressure relief valves protecting the vessel.
on LPG tanks by Moodie et al. (1988). In this study, a 5 tonne LPG
These tests can be compared to simulations performed using the
tank was fitted with two pressure relief valves with set pressures of
BLOWDOWN Technology in Aspen HYSYS, which is a general-
14.3 bar (gauge). Their operation was monitored to ensure that the
purpose solver for determining the dynamic behavior of the
pressure inside of the tank does not rise too high and cause a BLEVE
multi-component fluids contained in a pressurized vessel. It ac-
explosion. The tank pressure was measured with pressure trans-
counts for both the thermodynamic behavior of the contents of the
ducers at the top of the vessel where vapor is present and at the
vessel and the heat transfer from the environment to the vessel. It
bottom where liquid is present. The overall accuracy of the pressure
can also be used to model fire scenarios using both the fire equa-
measurements was ±0.1 bar. Tests were carried out with different
tions described in Eqs. (1) and (2).
liquid fill levels between 22% and 72% by volume. The tank was
When using Eq. (1) to model a fire scenario in BLOWDOWN, the
surrounded by three water tube calorimeters that were used to
heat is applied directly to the liquid inside the vessel. However, Eq.
measure the heat flux from the fire and thirteen thermocouples
(1) also represents the total heat absorbed from the fire and API 521
were used to measure the bulk fluid temperature inside the vessel
does not give any guidance on how this total absorbed heat should
on two vertical planes. The total duration of fire exposure was
be applied in a dynamic situation. In BLOWDOWN, it is assumed
30 min.
that the initial heat flux calculated by Eq. (1) with the initial wetted
Fig. 3 shows a flowsheet representation of the problem setup in
area remains constant over the duration of the dynamic simulation
BLOWDOWN. Because BLOWDOWN does not have a PRV but only a
as shown in Fig. 1.
constant-diameter orifice model, the results can only be compared
Alternatively, when using Eq. (2) to model a fire scenario in
until the time when the PRVs protecting the vessels open. To
BLOWDOWN, the heat flux is applied to the outer surface of the
simulate the period when the valves are closed, the orifice diameter
vessel wall. As shown schematically in Fig. 2, the heat flux is
initially the same in both the unwetted and wetted regions of the
wall. However, the vapor wall temperature increases much faster
than the liquid wall temperature because the heat transfer coeffi-
cient between the wall and the vapor inside of the vessel is very
low. Therefore, the heat flux to the unwetted region of the wall

Fig. 1. The heat flux applied to the liquid phase as a function of time when using the Fig. 3. Flowsheet representation of the LPG tank simulated with BLOWDOWN Tech-
empirical wetted area formula in API 521. nology in Aspen HYSYS.
 S. Biswas, B.J. Fischer / Journal of Loss Prevention in the Process Industries 50 (2017) 154e164 157

was set to zero. Moodie et al. (1988) mentions that fire usually shows the temperature versus time data for both the vapor and
engulfed the whole vessel after 100e150 s after ignition and that liquid phases. The temperature results are slightly higher and more
the heat flux absorbed by the vessel was insignificant before the conservative for both phases with the maximum deviation within
vessel was fully engulfed by fire. However, in the model, the fire 9 C for the liquid temperature and within 13 C for the vapor
flux must be applied from the beginning so the time is adjusted to temperature.
reflect the time after fire engulfment.
To model the fire, the experimental fire heat flux data reported
by Moodie et al. (1988) was averaged and then an initial heat flux 4. Depressurization case study: API wetted area equation VS.
was determined such that the same amount of total heat was added analytical equation
over the course of the simulation. In BLOWDOWN, the initial heat
flux at time zero can be directly specified such that the fire tem- In the previous section, simulations using BLOWDOWN Tech-
perature is back-calculated from Eq. (2). This fire temperature is nology in Aspen HYSYS were compared with experimental results
assumed to be constant over time during the simulation. Based on obtained from fire engulfment studies of LPG tanks. Using the API
the experimental data before the PSV opens for the first time, initial analytical equation for fire, both the pressure and vapor and liquid
heat fluxes of 81.22 kW/m2 and 75 kW/m2 were used for the 72% temperatures inside the vessel were accurately predicted. A more
liquid volume case and the 22% liquid volume case, respectively. complete study needs to investigate the effect of fire during a
Table 1 shows the parameters used for the analytical fire model for depressurization because this is often necessary when designing
both the 22% and 72% case. The parameters like fire emissivity, the size of a blowdown orifice. Because there is no experimental
surface emissivity, external heat transfer coefficient and hot gas data for a comparison, a case study is performed using the “S13”
ambient temperature were taken from the suggested default values case from Haque et al. (1992b) as BLOWDOWN has already been
for pool fire in API 521 Table A.3 (ANSI/API Standard 521 6th validated under ambient conditions for this situation. Table 3 pro-
Edition, 2014). Table 2 shows the geometrical parameters used in vides the detailed information for the vessel parameters and
the blowdown simulation and has been taken from Moodie et al. Table 4 gives detailed information about the scenarios investigated.
(1988). Fig. 6(a) shows the pressure profile with time for all four sce-
During the fire engulfment experiment, the fluid temperatures narios. As expected, the rates at which the pressures decrease are
were measured at thirteen fixed locations at different elevations on slightly lower for the fire scenarios when compared to the case of
two vertical planes. In the 72% liquid fill case, there is only one ambient conditions. In addition, for all fire scenarios, the pressure
location that consistently remained in the vapor zone. The top- profiles are similar initially. After approximately 600 s, the differ-
most location where the temperature was measured was initially ences in the pressure profiles increase but not significantly.
in the vapor zone. The temperature versus time curve for this Fig. 6(b) shows the liquid temperature inside the vessel for all
location shows that this temperature initially rises rapidly and then four scenarios. It can be seen that the liquid temperature does not
decreases to merge with temperature curves of other locations rise significantly during the simulation. This is due to the fact that
which are known to be in the liquid zone (Moodie et al., 1988). This most of the heat added to the liquid phase does not result in the
indicates that the liquid swells to this thermocouple level. The sensible heating of the liquid. Instead, the heat added to the liquid
BLOWDOWN simulation for this case also shows that the liquid is primarily used to provide the heat of vaporization to the liquid
level rises with time. Fig. 4 shows the internal pressure and liquid that evaporates into vapor. In the cases of the API wetted area
temperature as a function of time for the 72% liquid fill case. Both equation (Scenario #3) and the API analytical equation with initial
pressure and temperature are a good match with the experimental heat flux of 59.49 kW/m2 (Scenario #4), the liquid inventory of the
data. It should be noted that BLOWDOWN determines different vessel completely vaporizes around 1000 s. Because the contents of
temperatures for the vapor and liquid phases in the vessel; how- the vessel are only vapor, a liquid temperature is not reported after
ever, it does not determine a spatial distribution within these this time in the simulation.
phases but rather assumes a uniform temperature in each phase. The vapor temperatures inside of the vessel and downstream of
The maximum error for pressure is found to be less than 0.8 bar and the orifice are shown in Fig. 6(c) and (d). Using the wetted area
the maximum deviation in liquid temperature is less than 4C. equation does not predict an increase in temperature of the vapor
In the 22% liquid fill case, the top-most measurement locations or the temperature of the fluid downstream of the orifice. As dis-
for the temperature always remain in the vapor zone; therefore, the cussed, the heat applied to the liquid inventory inside the vessel
experimental temperature results for both the vapor and liquid primarily results in evaporation of liquid inventory. However, when
phases can be compared to the experimental measurements. using the analytical fire equation, heat is also transferred to the
Fig. 5(a) shows the pressure rise inside the vessel with time. In this vapor phase. As a result, the vapor temperatures increase signifi-
case, the BLOWDOWN results are found to be slightly higher and cantly, which could lead to a larger volumetric flow into the
more conservative with the maximum error within 2 bar. Fig. 5(b) disposal or flare network. Fig. 7 shows the volumetric flow through
the orifice when the flux calculated from the wetted area equation

Table 1
Input data for the 22% and 72% liquid fill by volume cases.

Parameters 72% Liquid case (by vol. %) 22% Liquid case (by vol. %)
2
Initial Heat Flux (kW/m ) 81.22 75
Fire Emissivity 0.75 0.75
Surface Emissivity 0.75 0.75
Vessel Absorptivity 0.75 0.75
Ambient fire temperature (C) 600 600
External Heat Transfer Coefficient (W/m2/C) 20 20
Fire Temperature (C) 942 914
158 S. Biswas, B.J. Fischer / Journal of Loss Prevention in the Process Industries 50 (2017) 154e164

Table 2
Input data for the 22% and 72% liquid fill by volume cases.

Parameters 72% Liquid case (by vol. %) 22% Liquid case (by vol. %)

Vessel Volume (m3) 10.25 10.25

Vessel Diameter (m) 1.68 1.68
Head Type Dished Dished
Cylinder Wall Thickness (mm) 11.85 11.85
Head Wall thickness (mm) 18 18
Composition LPG LPG
(C1: 0.06% (C1: 0.06%
C2:0.07% C2:0.07%
C3:94.96% C3:98.46%
C4:4.91%) C4:1.40%)
Orifice Diameter (mm) 0.0 0.0
Final Time (s)* 206 253

Fig. 5. Five tonne LPG tank (22% liquid by vol.): Evolution of the (a) vessel pressure and
Fig. 4. Five tonne LPG tank (72% liquid by vol.): Evolution of the (a) pressure and (b)
(b) temperature of the vapor and liquid phases inside the vessel after fire engulfment.
temperature of the liquid phase inside the vessel after fire engulfment.

vapor inventory is large enough to cause a noticeable change in the

without drainage or firefighting (Scenario #3) and when the same
vapor volumetric flow rate. Although in this case the differences are
flux was applied to the outside of the wall with the API analytical
marginal, this fact may be still be of some interest for safety engi-
equation (Scenario #4). It can be seen that the volumetric flow rates
neers responsible for design of flare networks and staggered
are slightly elevated at the later stages of simulation with the
blowdown studies.
analytical fire equation. This is mainly because the temperature of
Zamejc (2014) pointed out that one of the drawbacks of the API
the vapor inventory is the same at earlier times for both the wetted
analytical equation is that there are many possible permutations of
area equation and the analytical equation. With the analytical
input parameters like surface emissivity and external heat transfer
equation, the vapor gradually heats up with time. However, it is not
coefficient. Therefore, using the analytical equation runs the risk of
until the later stages of the simulation that the temperature of the
 S. Biswas, B.J. Fischer / Journal of Loss Prevention in the Process Industries 50 (2017) 154e164 159

Table 3 coefficient. For an open pool fire, the heat transfer coefficient is in
Vessel parameters used in the depressurization case study (Haque et al., the range of 10e30 W/m2/ C (ANSI/API Standard 521 6th Edition,
1992b).
2014, Table A.2). As above, the fire temperature is calculated such
Vessel Volume (m3) 2.73 that the initial absorbed heat flux is 45 kW/m2. Fig. 9 shows the
Vessel Diameter(m) 1.13 pressure profile and the vapor and liquid temperatures for different
Head Type Dished
Cylinder Wall Thickness (mm) 59
external convective fluxes of 10, 20, and 30 W/m2/ C.
Head Wall thickness (mm) 50 It can be seen that if the initial absorbed flux is fixed, the profiles
Composition C1: 66.5% for the pressure and fluid temperatures are virtually the same even
C2: 3.5% though the emissivity and heat transfer vary. In a vessel containing
C3: 30%
hydrocarbon liquid, the liquid inventory will absorb most of the
Liquid Amount 20%
Final Time (s) 1500 heat from the fire, which causes the temperatures of the liquid and
Orifice size(mm) 10 the wetted wall to increase. However, these temperatures will
never reach a temperature higher than the bubble point of the
hydrocarbon liquid. As a result, the decrease in heat flux to the wall
being wrongly applied, which has been an impediment for a wide- at a later time, t, (i.e. Dqabsorbed ) is very low compared to the initial
adoption of this method among safety engineers. Understanding heat flux (qabsorbed jtime¼0 ) for hydrocarbon liquids. Therefore,
the effect of various parameters can help guide the use of the neither the surface emissivity nor the external heat transfer coef-
analytical equation. ficient has significant effects on either the pressure profile or the
Experimental studies of fire engulfment have shown that fire temperature in the case of the vessel containing liquid. As will be
temperature remains relatively constant with time for the duration shown in the next section, this is not necessarily true in unwetted
of the fire (Anderson et al., 1974; Moodie et al., 1985, 1988). In most situations where the wall temperatures can get significantly higher.
cases, wall temperature is the only parameter that is a function of
time in the Eq. (2). As a result, the fire temperature can be calcu-
5. Wall temperature analysis of a vessel with piping using a
lated from the initial absorbed heat flux as was done earlier when
local heat flux
performing the comparison with experimental results. Further-
more, the decrease in the heat flux to the system at later time, t, can
Dynamic simulation can be used to determine the required size
be written as:
of a blowdown orifice by analyzing the pressure response in a
vessel under fire. However, in some hydrocarbon processing facil-
Dqabsorbed ¼ qabsorbed jtime¼0  qabsorbed
ities (particularly offshore) there is a limitation in flare capacity,
  which imposes a constraint on freely sizing the blowdown orifice.
Dqabsorbed ¼ sεs Ts ð0Þ4  Ts ðtÞ4 þ hðTs ð0Þ  Ts ðtÞÞ (3) For such cases, providing passive fire protection of critical process
segments is suggested to protect the vessels and other process
Because the absorbed heat flux (qabsorbed ) is the boundary con- piping segments from failure (Salater et al., 2002). In a real fire case,
dition for the system consisting of the vessel and the liquid and the heat load from the fire to the vessel or process equipment is
vapor inventory inside the vessel, the pressure and temperature non-uniform. The pressure profile inside the vessel depends on the
profile of the vessel inventory with time only depends this condi- average heat load in such a case. However, the process equipment
tion. From Eq. (3), it can be seen that the decrease of this absorbed and piping can experience significant thermal weakening in an
heat flux at a later time (t) only depends on the surface emissivity isolated spot due to a local peak heat flux, which can result in
(εs ) and the external heat transfer coefficient (h). This implies that if mechanical failure (Salater et al., 2002; ANSI/API Standard 521 6th
the initial heat flux (qabsorbed jtime¼0 ) is fixed, then the absorbed heat Edition, 2014). To capture the effect of thermal weakening in a
flux boundary condition, and therefore the dynamic response of the simulation, a local peak heat flux is applied to a specific location
vessel and its inventory will also depend only on these two that is a small fraction of the total area exposed to the fire. Then, the
parameters. transient behavior of the wall temperature at this location can be
According to the API 521 guidance, the range of surface emis- obtained. This can be used to determine whether the temperature
sivity is between 0.3 and 0.8 (ANSI/API Standard 521 6th Edition, rise is enough to weaken the wall material substantially and cause
2014, Table A.2). Using the value of surface emissivity and the rupture. In a real fire, there is no guarantee that the peak heat flux
default value for open fire of 45 kW/m2 for the initial absorbed heat impacts just one location during the whole depressurization.
flux (ANSI/API Standard 521 6th Edition, 2014, Table A.3), the fire However, this is the most conservative way of designing for fire
temperature can be calculated from Eq. (2). Fig. 8 shows the pres- cases.
sure profile and the vapor and liquid temperature profiles for To apply this approach, it is necessary to study the detailed
different surface emissivity values of 0.3, 0.5, and 0.8 for the S13 geometry of the process equipment under fire instead of a lumped
case defined previously. volume. Fig. 10 shows a schematic diagram of the system investi-
The convective heat transfer from the external hot gases sur- gated in which a vessel is connected by a pipe to the blowdown
rounding the vessel is determined using the external heat transfer valve. It is also connected with a blocked outlet valve through

Table 4
Fire scenarios applied to S13 case.

Scenario #1 Scenario #2 Scenario #3 Scenario #4

No External API analytical equation with initial Heat Flux directly to the liquid ¼ 59.49 kW/m2 (This value API analytical equation with initial heat flux to vessel
Fire Flux heat flux to vessel wall ¼ 45 kW/m2 is calculated from API wetted area equation without wall ¼ 59.49 kW/m2 (Flux obtained from API wetted area
(API default for Open Fires, Table A.3) drainage and firefighting. BLOWDOWN, assumes that the equation was applied to wall instead of applying to the
initial heat flux calculated initial wetted area remains fluid directly)
constant over the duration of the dynamic simulation.)
160 S. Biswas, B.J. Fischer / Journal of Loss Prevention in the Process Industries 50 (2017) 154e164

Fig. 6. S13 depressurization case (20% liquid inventory by vol.): Evolution of the (a) vessel pressure, (b) liquid temperature, (c) vapor temperature, and (d) fluid temperature at the
orifice outlet with time for different fire loads.

average initial heat flux of 45 kW/m2 is applied but a peak initial

flux of 120 kW/m2 is applied on 0.1% of the area. Table 5 summa-
rizes the dimensions in each of the individual vessel and pipes.
Fig. 11 shows the pressure profiles for the case of the vessel with
piping and the case of a single vessel. Both systems depressurize at
the same rate because the volume of the pipes are relatively small.
Fig. 12 shows the vapor zone wall temperature of the vessel and
pipes. Because the pipe walls are not as thick as the vessel wall, the
vapor zones of the pipes heat up much faster than the vapor zone of
a vessel. Additionally, as expected, the temperature rises in the
pipes exposed to the peak flux of 120 kW/m2 are much more rapid
than the temperature rise of the vessel exposed to the average flux
of 45 Kw/m2. However, the pipe connected to the blowdown orifice
heats up much less than the pipe connected with the blocked vapor
outlet even though both pipes were exposed to peak flux of
120 kW/m2. This is due to the fact that the gas velocity in the pipe
Fig. 7. S13 depressurization case (20% liquid inventory by vol.): Volume flow through connected to the blowdown orifice is much higher than the gas
the orifice as a function time for the API 521 wetted area and analytical fire models. velocity in the pipe connected to the blocked vapor outlet. This
results in higher internal convective heat transfer to the vapor
going out of the orifice. Hence, the vapor wall of this pipe does not
another pipe, which remains closed during the depressurization. heat up as rapidly as that of the pipe connected to the blocked
The process conditions and inventory is the same as in our previous vapor outlet. Another observation that can be made from Fig. 12 is
study given in Table 3. Initially, the pipes are filled completely with that for the pipe connected with the blocked vapor outlet, the peak
vapor. wall temperature flattens out after 400 s. In this region, the heat
In this case study, the average heat flux of 45 kW/m2 is applied absorbed by the vessel wall from the external fire is the same as the
to the vessel. However, the total heat flux applied to the blowdown heat transfer to the inventory of the vessel.
and vapor outlet pipes is treated differently. In the pipes, a global
 S. Biswas, B.J. Fischer / Journal of Loss Prevention in the Process Industries 50 (2017) 154e164 161

Fig. 9. S13 depressurization case (20% liquid inventory by vol.): (a) Vessel pressure and
Fig. 8. S13 depressurization case (20% liquid inventory by vol.): (a) Vessel pressure and (b) liquid and vapor temperatures as a function of time for different external heat
(b) liquid and vapor temperatures as a function of time for different surface emissiv- transfer coefficients with initial fire heat flux of 45 kW/m2.
ities with an initial fire heat flux of 45 kW/m2.

the initial period less than 200 s for three different heat transfer
Although not performed in this study, Salater et al. (2002) and coefficients (10, 20, and 30 W/m2/C).
Scandpower Risk Management AS (2004) discuss how the wall As previously shown, for a single vessel, neither the surface
temperature data obtained using the peak flux can be used to emissivity nor the external heat transfer coefficient has strong ef-
determine material failure. As the wall temperature in the vapor fects on the pressure profile, vapor and liquid temperature profiles,
zone rises, there is thermal weakening of the vessel due to a loss of and wall temperatures. However, in the associated piping with a
ultimate tensile strength (UTS) of the material. If the temperature- local peak heat flux, the effect of surface emissivity on wall tem-
dependent data of the UTS of the material is available, then the UTS perature can be significant. If the surface emissivity is varied, then
can be compared with the actual calculated stress in the pipe to Eq. (3) shows that the reduction of the heat flux will vary as the 4th
verify when this mechanical failure will occur. Scandpower Risk power of wall temperature. However, if the external heat transfer
Management AS (2004) provide details on the required safety coefficient is varied, then Eq. (3) shows that the reduction of the
factors, wear and tear, and other mechanical loading effects that heat flux will vary linearly with wall temperature. Fig. 14 shows the
need to be considered when embarking on this type of analysis. decrease of the heat flux as a percentage of the initial peak flux
As in the previous section, the effects of the surface emissivity (120 kW/m2) with the vessel wall temperature for different surface
and external heat transfer coefficients on the wall temperature emissivity values and for different external heat transfer co-
profiles of the pipes that are exposed to peak flux are investigated. efficients. As shown, the differences in the heat flux are not sig-
The fire temperature is calculated from Eq. (2) such that the initial nificant until the wall temperature reaches approximately
local peak heat flux is 120 kW/m2. Fig. 13(a) shows the wall tem- 500e600 C. In a vessel containing hydrocarbon liquid, the liquid
perature profiles in the vapor zone for the blocked vapor outlet pipe inventory or the wetted wall will never reach this temperature
exposed to the peak flux for three different values of the surface because it is significantly higher than the bubble point of typical
emissivity (0.3, 0.5 and 0.8). As the surface emissivity decreases, the hydrocarbon liquids. Because the liquid inside of the vessel will
final wall temperature increases significantly. Conversely, varying absorb most of the heat from the fire, varying surface emissivity has
the heat transfer coefficient has a smaller impact on the final wall no effect on the pressure profile, vapor and liquid temperatures,
temperature. As shown in Fig. 13(b), the wall temperature profiles in and wall temperatures. However, in the portion of the vapor outlet
the vapor zone do not significantly change and are nearly identical in pipe where the peak heat flux of 120 kW/m2 is applied, the wall
162 S. Biswas, B.J. Fischer / Journal of Loss Prevention in the Process Industries 50 (2017) 154e164

Fig. 10. Schematic diagram of the depressurization system where the vessel is connected to pipes in the outlet line and the blowdown line.

Table 5
Dimensions of individual vessel and pipes.

Parameters Vessel Pipe (Blocked Outlet) Pipe (Blowdown Line)

3
Vessel Volume (m ) 2.73
Diameter 1.13 m 100 mm 100 mm
Pipe Length 5m 5m
Head Type Dished e e
Cylindrical Wall Thickness (mm) 59 mm 10 mm 10 mm
Head Wall thickness (mm) 50 e e
Initial Liquid Amount 20% 0% 0%
Area Exposed to Global Average Heat Flux (45 kW/m2) 100% 99.9% 99.9%
Area Exposed to Peak Average Heat Flux (1202) 0% 0.1% 0.1%
Composition C1: 66.5%
C2: 3.5%
C3: 30%
Final Time (s) 1500
Orifice size(mm) 10

Fig. 12. S13 depressurization case (20% liquid inventory by vol.): Wall temperature at
Fig. 11. Effect of the connected piping on the pressure in the system. the location of the peak flux (120 kW/m2) as a function of time.
 S. Biswas, B.J. Fischer / Journal of Loss Prevention in the Process Industries 50 (2017) 154e164 163

Fig. 14. Percentage reduction in the heat flux as a function of wall temperature for
different (a) surface emissivities and (b) external heat transfer coefficients for an initial
heat flux of 120 kW/m2.
Fig. 13. Wall temperature at the location of the peak heat flux in the blocked vapor
outlet line as a function of time for different (a) surface emissivities and (b) external
heat transfer coefficients with an initial absorbed heat flux of 120 kW/m2.
that are difficult to obtain, it was shown that only the surface
emissivity and heat transfer coefficient affect the absorbed heat
temperature goes well beyond 500e600C and as a result the wall flux with time. In a vessel undergoing depressurization, the surface
temperature increase significantly. Lower emissivity leads to lower emissivity does not affect the predictions for either the pressure or
re-emitted heat flux from the vessel, and therefore higher absorbed fluid phase (vapor and liquid) temperatures inside of the vessel.
heat flux by the equipment wall. This leads to a higher and thus This is because the wall temperature never gets high enough to
more conservative wall temperature predictions. cause a significant difference in the absorbed heat flux. However, it
was shown that the wall temperature in a vapor-filled pipe exposed
6. Conclusion to a local peak heat flux will increase significantly with a decrease
in surface emissivity. In addition, the flow in the vapor-filled pipe
Predictions using the analytical fire equation from API 521 with affects the wall temperature significantly because of the enhanced
the BLOWDOWN Technology in Aspen HYSYS compared well with heat transfer between the flowing vapor and the pipe wall. As a
available experimental data from fire engulfment studies of LPG result, in depressurization systems, the wall temperature of pipes
tanks. By using the proper amount of heat flux, the pressure profiles with significant flow, such as the blowdown line, do not get as high
and vapor and liquid temperature profiles were consistent with as that of pipes connected to blocked outlets with lower flows. In
experimental data from Moodie et al. (1988). In addition, results both the vessel and pipe, varying the external heat transfer coef-
from comparison studies of different fire models show that the ficient was found to exert an insignificant effect. From these ob-
pressure profiles using the empirical wetted area equation com- servations, it can be concluded that the absorbed heat flux at the
pares well with the analytical equation provided that the initial initial time is the key factor when sizing an orifice in a system
heat flux used in the analytical fire model is the same as the heat exposed to fire. But when extending the analysis to material failure
flux directly applied to fluid in the wetted area model. However, the using a local peak heat flux, lower surface emissivity values lead to
wetted area model does not account for the sensible heating of the more conservative results for wall temperatures.
vapor and the resultant temperature increase of the fluid down- When analyzing a fire scenario, safety engineers often design a
stream of the orifice. This leads to marginally lower volumetric flow blowdown valve by determining the orifice size required to reduce
predictions in the downstream flare networks. the pressure to 50% of the design pressure within 15 min. By using
Although the API 521 analytical equation has several parameters the tools and methods in this study, wall temperatures in the vessel
164 S. Biswas, B.J. Fischer / Journal of Loss Prevention in the Process Industries 50 (2017) 154e164

and piping can be calculated more accurately during blowdown Systems.

Balke, C., Heller, W., Konersmann, R., Ludwig, J., 1999. Study of the Failure Limits of a
and therefore the orifice size can be based on a stress analysis
Railway Tank Car Filled with Liquefied Petroleum Gas Subjected to an Open
rather than the traditional “15 min rule”. Evaluating the pressure Pool Fire Test. BAM project, p. 3215.
and thermal stress profiles with a comparison to the yield strength Haque, M.A., Richardson, S.M., Saville, G., 1992a. Blowdown of pressure vessels. I.
of the material will allow engineers to determine the minimum Computer model. Trans. IChemE, Part B, Proc. Safe Env. Prot. 70 (B1), 3e9.
Haque, M.A., Richardson, S.M., Saville, G., Chamberlain, G., Shirvill, L., 1992b.
blowdown rate that can avoid rupture. It should also be pointed out Blowdown of pressure vessels. II. Experimental validation of computer model
that this study still assumes that the vessel is engulfed by the fire. and case studies. Trans. IChemE, Part B, Proc. Safe Env. Prot. 70 (B1), 10e17.
Further investigations are required to estimate the heat flux in the Ludwig, J., Heller, W., 1999. Fire Test with a Propane Tank Car. BAM Test Report III.2/
9907.
Stefan-Boltzmann equations for a fire distant to the process Moodie, K., Billinge, K., Cutler, D.P., April 1985. The fire engulfment of LPG. storage
equipment, which would require a full three-dimensional study of tanks'. I. Chem. E. Symp. Assess. Control Major Hazards. DMIST.
the heat transfer inside the process equipment. Moodie, K., Cowley, L.T., Denny, R.B., Small, L.M., Williams, I., 1988. Fire engulfment
tests on a 5 tonne LPG tank. J. Hazard. Mater. 20, 55e71.
Mahgerefteh, H., Falope, G.B., Oke, A.O., 2002. Modeling blowdown of cylindrical
Acknowledgements vessels under fire attack. AIChE J. 48 (2), 401e410.
Melhem, G.A., Gaydos, D., 2015. Properly calculate vessel and piping wall temper-
atures during depressuring and relief. Process Saf. Prog. 34 (1), 64e71.
The authors are grateful to Dr. Graham Saville and Professor Roberts, T.A., Buckland, I., Shirvill, L.C., Lowesmith, B.J., Salater, P., 2004. Design and
Stephen Richardson, formerly of Imperial College London, for protection of pressure systems to withstand severe fires. Process Saf. Environ.
invaluable discussions. Prot. 82 (2), 89e96.
Salater, P., Overaa, S.J., Kjensjord, E., 2002. Size depressurization and relief devices
for pressurized segments exposed to fire. Chem. Eng. Prog. 98 (9), 38e45.
References Scandpower Risk Management AS, 2004. Guidelines for Protection of Pressurized
Systems Exposed to Fire. Scandpower, AS, Norway.
Anderson, C., Townsend, W., Zook, J., Cowgill, G., 1974. The Effects of a Fire Envi- Zamejc, E., 2014. API Standard 521 new alternative method to evaluate fire relief for
ronment on a Rail Tank Car Filled with LPG. Rept No FRA-OR&D. US Dept of pressure relief device sizing and depressuring system design. J. Loss Prev.
Trans, p. 75, 31. Process Industries 27, 21e31.
ANSI/API Standard 521 6th Edition, 2014. Pressure-Relieving and Depressuring