Computers and Chemical Engineering 25 (2001) 793798

Dynamic simulation of the behaviour of pressure relief systems

Juha-Pekka Pokki *, Markku Hurme, Juhani Aittamaa
Helsinki Uni6ersity of Technology, Laboratory of Chemical Engineering and Plant Design, P.O. Box 6100, FIN-02015 HUT, Finland
Received 10 May 2000; accepted 5 January 2001
Abstract
Pressure relief of near critical processes working at two phase region is simulated dynamically. The design procedure for
pressure relief devices is demonstrated by an example. The example shows that the required relieving area of the safety valve
depends on time and scenario. This is the basis for the selection of the type and the size of safety valve. When the process relieves
through the selected valve the outow and the pressure of the protected vessel varies dynamically. This phenomenon cannot be
discovered by the static approach that is normally used in the design of pressure relieving devices. The traditional sizing equations,
API 520 and DIERS, are also difcult to apply in near critical region and they include coefcients that are sometimes difcult
to obtain. The sensitivity may be so large that a different size of the relieving device is selected depending on the model. Usage
of two common models, API 520 and DIERS, in the region of two phases is discussed. 2001 Elsevier Science Ltd. All rights
reserved.
Keywords: Dynamic simulation; Modelling; Relief system
www.elsevier.com/locate/compchemeng
1. Introduction
In a chemical process disturbances may take place
which cannot be handled by the control system. These
disturbances include loss of cooling, power failure,
blockage of process equipment or piping, malfunction
of control equipment, feed at abnormal rate, abnormal
weather conditions, etc. As a consequence an increase
in pressure of some process equipment may result. The
equipment is designed to tolerate a certain maximum
pressure that should not be exceeded because that may
cause the equipment to rupture. Pressure relieving
devices prevent excessive pressure and operate indepen-
dently of the control system. The devices open a route
for the material to relieve into a safe location. Often the
design of pressure reliving devices is based on an as-
sumption of stationary behaviour of the system. How-
ever, the behaviour of such systems is nearly always
dynamic. During the relief the uid relieved may be in
liquid, vapourliquid mixture, gas or supercritical
phase. This gives challenges for the design of safety
systems that must be capable to handle various process
conditions. In many cases the answer to these problems
is dynamic simulation which provides more information
on the behaviour of the pressure relief systems than the
stationary approach does. The most common design
methods are API 520, (American Petroleum Institute,
March, 1993), and DIERS, (Leung, 1996). In these
methods the design problem is divided into separate
regions depending on the phase of the relieving uid. It
is essential that the simulator identies the phase, pro-
vides the physical properties of the relieving uid and
shifts smoothly from one region to another when the
phase of the relieving uid changes. In this paper API
520 and DIERS models are compared. The simulation
reveals that there might be large deviations depending
on the model used. The selection of model and scenario
is important in designing pressure relief systems.
2. Principles of the modelling
In the previous paper, (Pokki, Aittamaa, Keskinen,
& Hurme, 1999), a dynamic simulator for calculating
abnormal process conditions of chemical processes was
presented. The simulator operates in two modes, design
and rating mode. The simulation in the design mode
gives the required outow of uid, polymer and catalyst * Corresponding author.
0098-1354/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S0098- 1354( 01) 00666- 4
J.-P. Pokki et al. / Computers and Chemical Engineering 25 (2001) 793798 794
and the discharge area to keep the reactor pressure
equal to the set pressure of the pressure relieving device
at every time step. The phase of the relieving uid
depends on the location of the pressure relieving device
in the reactor, usually pressure relieving devices are
placed at the top of the reactor or vessel. The polymer
particles ow out with the uid if the polymer does not
settle down inside the reactor once the abnormal event
starts. That is why this simulator allows also the calcu-
lation of polymer ow in the design mode. The uid in
this work means hydrocarbons and also quantum gases
like hydrogen and nitrogen in liquid, vapour, gas, va-
pourliquid or supercritical state. In the rating mode
the pressure relieving device is specied and tempera-
ture, pressure and many other thermodynamic proper-
ties of the reactor are calculated at every time step. This
mode requires the specication of the pressure relieving
device. It includes its discharge area, valve coefcients
and the characteristic curve, i.e. the opening of the
valve as a function of pressure difference.
The uid owing through the relieving device may
split into two phases. In the previous paper this was not
studied because the mass ow was the primary interest.
The basic equations presented in the previous paper,
(Pokki et al., 1999), are not repeated here. To sum-
marise the model it was assumed that the reactor is a
vessel that has a dened volume. The shape of the
reactor was not considered. If the reactor is tall the
pressure in the bottom is higher due to hydrostatic
pressure and there may be also temperature gradients.
The heat capacity of the wall of the vessel and the effect
of solved polymer in the uid are omitted. The reaction
rate of each monomer depends on temperature and on
concentration of monomer and catalyst. It is assumed
that the ow in inlet and outlet pipe does not affect the
pressure relieving device. In the previous paper, (Pokki
et al.), the speed of sound was used as the speed of
owing uid. The validity of this assumption decreases
when the uid becomes dense. That is why the kinetic
energy term is removed from the energy balance.
3. Valve models
API 520, (American Petroleum Institute, March,
1993), is a widely used and it is a recommended practice
in petroleum industry. The equations are divided into
vapour/gas equations and liquid equation. The selec-
tion between critical or subcritical vapour/gas ow is
based on critical ow pressure p
cf
.
p
cf
p
1
=
2
k+1
n
k/(k1)
, (1)
where k=c
p
/c
6
is ratio of heat capacities and p
1
is the
inlet pressure. When p
cf
is greater than back pressure,
p
b
, the vapour/gas ow is critical and the API 520
equation for orice area A is as follows
A
in
2
=
W/(1b/h)
CK
d
(p
1
/psia)K
b
' (T/R)Z
M/(g/mol)
, (2)
and when p
cf
Bp
b
the vapour/gas ow is subcritical the
equation is as follows
A
in
2
=
W/(1b/h)
735F
2
K
d
' (T/R)Z
M/(g/mol)(p
1
/psia)[(p
1
/psia) (p
2
/psia)]
.
(3)
When the relieved uid is liquid the sizing equation is
A
in
2
=
W/gal.US/min)
38K
d
K
w
K
6
' G
(p
1
/psig)(p
2
/psig)
. (4)
It is assumed in Eq. (4) that the uid remains liquid as
it ows through the valve. The variables and values of
coefcients K
d
, K
b
, F
2
, K
w
, and K
6
, are given in API 520.
When these Eqs. (1) (4), are used in near critical
conditions some questions arise. The critical pressure
ratio is dependent on ratio of heat capacities. The heat
capacity at constant pressure c
p
increases when the
critical point is approached and the value of k can be
much larger than 2 that is the highest value mentioned
in API 520. Also in the near critical conditions the
value of compressibility factor deviates signicantly
from unity. Two phase ow is handled by combining
these equations. This procedure is not clearly specied
in API 520. The uid is ashed either to critical ow
pressure or to back pressure whichever is greater. Then
the area for vapour and liquid is computed with va-
pour/gas and liquid equations, respectively. The re-
quired area of the valve is equal to the sum of vapour
and liquid areas. The selected area is equal to or greater
than the required area. According to Leung and
Nazario (1990) the API 520 method may lead to under-
sizing of pressure relieving devices. They modied the
API 520 liquid equation by using the pressure drop
from upstream pressure to critical ow pressure instead
of pressure drop from upstream pressure to down-
stream pressure. This modication gave better results in
their work.
If the vessel is exposed to re the sizing of pressure
reliving device requires the rate of heat transfer to the
contents of the vessel. If the vessel is lled with liquid
the so-called wetted surface method of API 520 is
commonly used. The heat transferred is
Q
BTU/h
=21000F
A
ft
2

0.82
(5)
where A is the effective wetted surface area and F is the
environment factor. Environment factors are tabulated
in API 520. They are sensitive to engineering judge-
ment, since the insulation may deteriorate in time and
loose its re protection capability. In that case the heat
transferred from ames to vessel is remarkably higher
J.-P. Pokki et al. / Computers and Chemical Engineering 25 (2001) 793798 795
and may lead to excessive overpressure. Another impor-
tant aspect is the wetted area in cases when the vessel
is only partially lled with liquid or is supercritical. The
wetted surface usually decreases as the relief proceeds
reducing the heat rate. It seems that the assumption of
constant heat rate based on initial condition leads to a
conservative design. The problem with the supercritical
uid is that the uid cannot be characterised clearly as
liquid or vapour/gas for Eqs. (1) (4) since the uid has
both liquid and vapour like properties at the same time.
The DIERS method, (Leung, 1996), for two phase
ow is more precisely dened than API 520. It has been
claimed to apply for systems where the reduced temper-
ature is less than 0.9 and the reduced pressure is less
than 0.5. Above those conditions it underestimates the
mass ux and thus overpredicts the area. The method is
developed primarily for one-component ashing uids.
The calculation procedure of DIERS (Leung, 1996) and
(Nazario & Leung, 1992) determines rst the -
parameter.
=
x
0
6
g0
6
0
+
C
f0
T
0
p
0
6
0
6
fg0
h
fg0

2
, (6)
where x
0
is mass fraction of vapour at inlet, C
f0
is
specic heat capacity of liquid at inlet, T
0
temperature
at inlet, p
0
is pressure at inlet. Subscript fg refers to
liquidvapour transition, g refers to vapour. Specic
volume of uid w
0
at inlet conditions is
6
0
=x
0
6
g0
+(1x
0
)6
f0
. (7)
Variables are explained in (Leung, 1996) and
(Nazario & Leung, 1992). The critical pressure ratio p
c
p
c
=
p
c
p
0
, (8)
is computed, as -parameter is known, from
p
c
2
+(
2
2)(1p
c
)
2
+2
2
ln(p
c
) +2
2
(1p
c
) =0.
(9)
This leads to iterative solution of critical pressure
ratio. In this work the explicit expression for critical
pressure ratio is obtained from Nazario and Leung
(1992) and used as an initial guess for iteration.
p
c
=0.55+0.217(ln ) 0.046(ln )
2
+0.004(ln )
3
.
(10)
The Eq. (10) is good for initial guess if \0.4.
When 0BB0.4 the initial guess is p
c
=0.3. If the
relieving uid is close to the thermodynamic critical
point, the -parameter can be determined from, (Le-
ung, 1996),
=
6
1
6
0
1
,p
0
p
1
1

. (11)
The subscript 1 refers to the properties determined at
70% of inlet pressure via isenthalpic ash. It is also
suggested by (Leung, 1996) to determine -parameter
for non-ideal systems at 90% of inlet pressure via
isenthalpic ash. After the critical pressure ratio is
solved the mass ux G in orice is calculated from
G
(p
0
/6
0
)
1/2
=
'
2

ln
p
p
0

+(1)

1
p
p
0

p
0
p
1

+1

, (12)
when critical ow pressure p
c
is less than back pressure
p
b
the ow is not choked and p is set equal to p
b
. If
p
c
\p
b
the ow is choked and mass ux G is computed
from
G
(p
0
/6
0
)
1/2
=
p
c

1/2
. (13)
If the ow in DIERS model, (Leung, 1996), enters
gas region the mass ux is computed from
G
p
0
z
g0
=0.606. (14)
where z is density. The area of the orice is
A=
W
KG
, (15)
where W is required mass ow and K is the valve
discharge coefcient.
4. Example
This example consists of a reactor, volume 30 m
3
,
two storage tanks, both 70 m
3
and a buffer tank, 120
m
3
, see Fig. 1.
The initial molar composition of the reactor, tank 1,
tank 2 and the buffer in mole fractions is; hydrogen
0.01, propane 0.10, propene 0.80, n-butane 0.05 and
n-pentane 0.04. The computed thermodynamic critical
point is 376 K and 4.9 MPa abs, based on the method
of Heidemann and Khalil (1980). The amount of poly-
mer in the reactor is 15 wt%. It is assumed that the
reactor is at the two phase region when the runaway
begins. The reaction rate in the reactor is temperature Fig. 1. Schematic gure of the system.
J.-P. Pokki et al. / Computers and Chemical Engineering 25 (2001) 793798 796
Table 1
Required valve area from different models simulated in design mode
a
Reactor 2000 l 1000 DIERS s2p (Ds2p)
mm
2
API 520 (A2p) 1500 l 750 mm
2
2500 l 1400 mm
2
DIERS 07ad (D07ad)
Mod API 520 (Am2p) 1800 l 800 mm
2
2600 l 1000 DIERS s3p (Ds2p) Tank 1
mm
2
API 520 (A2p) 2400 l 1100 mm
2
3000 l 1200 mm
2
Mod API 520 (Am2p)
Tank 2 1300 l 1100 mm
2
DIERS s2p (Ds2p);
1050 l 950 mm
2
API 520 (A2p);
1500 l 1300 mm
2
Mod API 520 (Am2p);
a
The required valve area varies between shown limits during
simulation.
surface decreases as the uid is relieved. The content of
the buffer at 303.15 K is in vapour region. The purpose
of the buffer tank is to remove the peaks of the ow to
the are and partly store the vented material.
This example demonstrates the sensitivity of the dy-
namic simulation of the pressure reliving systems to
different models and assumptions. The system is simu-
lated in the design mode and the results are presented
in Table 1. The abbreviations are; (Ds2p) refers to
saturated two-phase ow of DIERS, (D07ad) to near-
critical two-phase ow of DIERS, (A2p) to two-phase
ow of API 520, (Am2p) to modied two-phase ow of
API 520 model by Leung and Nazario (1990). The
discharge coefcient of DIERS model is set to 0.81.
The equation of state used is PengRobinson (Peng &
Robinson, 1976).
The required area of safety valve, (SV), of the reactor
is greatest if the near-critical two-phase DIERS model
(D07ad) is used. The API 520 model (A2p) gives the
smallest required area. The required area is given as a
function of time and the selected area of SV is selected
from API 526, (American Petroleum Institute, June,
1995). In this case K, 1185 mm
2
is selected because it
is a compromise between results of API 520 and
DIERS models. There is no polymer ow out in this
example but the calculation of polymer containing relief
is discussed by (Pokki et al., 2000). The tanks 1 and 2
are identical in volume, content and initial state but the
amount of heat rate from ames to the tank 2 is only
half of the heat rate to tank 1. This situation demon-
strates the difculties in estimating the insulation fac-
tor. The selected size of safety valves of tanks 1 and 2
is M, 2322 mm
2
which is also a compromise between
API 520 and DIERS models. The set pressures of tanks
are 1.5 MPa abs.
The sensitivity of relief valve sizing is studied by
simulating the system using different API 520 and
DIERS combinations with the sizes of SVs selected
based on the design mode. The vapour/gas ow from
buffer tank to are starts as pressure increases over 0.2
MPa abs. The ow area is set to 10332 mm
2
to keep its
pressure below 1 MPa abs. The ow is modelled with
API 520 vapour/gas equations. The models used in the
simulations are presented in Table 2.
The simulated pressure of reactor is shown in Fig. 2.
The set pressure of SV of reactor is 4.0 MPa abs. The
modied API model gives the maximum pressure below
+10% overpressure which is acceptable but both
DIERS models give overpressure higher than 10%.
The mass ow from reactor is shown in Fig. 3. The
sensitivity in mass ows computed from the models is
so large that it affects the design of the discharge pipe.
The discharge pipes should be designed conservatively.
The pressure of tank 1 exposed to re varies as
shown in Fig. 4. Pressure starts to increase faster when
the DIERS model than API models are applied. The
increase is in accepted limits.
Table 2
Combinations of valve models used in the simulations in rating mode
DIERS s2p (Ds2p) Reactor Case a
DIERS s2p (Ds2p) Tank 1
Tank 2 API 520 (A2p)
Case b Reactor DIERS 07ad (D07ad)
API 520 (A2p) Tank 1
Tank 2 DIERS s2p (Ds2p)
Reactor Mod API 520 (Am2p) Case c
Tank 1 Mod API 520 (Am2p)
Tank 2 Mod API 520 (Am2p)
Fig. 2. Simulated pressure of reactor in rating mode.
dependent and follows the Gaussian distribution
shaped curve. Reaction rate is 1.86 kg/s at initial tem-
perature of reactor 352.15 K and 2.22 kg/s at 373 K
that is the maximum reaction rate. The contents of the
storage tanks are at 298.15 K in two phase region. Both
storage tanks are nearly full of liquid and the relief is
two phase ow. Both tanks are exposed to re and the
heat transferred is constant although the area of wetted
J.-P. Pokki et al. / Computers and Chemical Engineering 25 (2001) 793798 797
The mass ow from tank 1 is shown in Fig. 5. There
are large differences in the time when the ow stops at
the rst time.
Fig. 6. Simulated pressure of tank 2 in rating mode.
Fig. 3. Simulated mass ow from reactor in rating mode.
Fig. 7. Simulated mass ow from tank 2 in rating mode.
Fig. 4. Simulated pressure of tank 1 in rating mode.
The simulated pressure of tank 2 is in Fig. 6. The
heat transferred from ames to the tank 2 is only half
of the heat of tank 1 and thus the rise of pressure in
tank 2 is much slower. Because the ow of heat is
smaller there is no accumulation in pressure.
The mass ows from tank 2 show substantial devia-
tions between models as seen in Fig. 7 because the
smaller mass ow causes longer duration of relief.
Finally, the simulated pressure of buffer is shown in
Fig. 8. The start of relief from reactor (88 s), tank 1
(322 s) and tank 2 (644 s) can be seen as rapid increase
in pressure in Fig. 8.
The relief from buffer is in Fig. 9. The peak ow is
nearly equal in all models but it happens at different
time. The ow stays between 10 and 20 kg/s most of the
time, because the reactor relieves before the tanks 1 and
2 start to relieve. Sometimes the capacity of the are
may be the limiting factor in plant design. The dynamic
simulation helps in adjusting the volume of the buffer
tank and the ow to are. Usually increasing volume of Fig. 5. Simulated mass ow from tank 1 in rating mode.
J.-P. Pokki et al. / Computers and Chemical Engineering 25 (2001) 793798 798
Fig. 8. Simulated pressure of buffer in rating mode.
5. Conclusion
It was shown that the results of the relief valve
simulation are sensitive to the models used. The selec-
tion of model is cumbersome because these are specic
to certain regions and the process may enter various
regions during emergency relief. The models suffer also
experimental validation at near critical conditions. In
addition the selection of the scenario is very important
in design to nd out the biggest relief ow and the
highest peak pressure. The near critical conditions are a
challenge for simulation and modelling. The consider-
ation of the dynamic nature of the system is crucial in
design of pressure relieving devices. The dynamics re-
veals the behaviour of buffer tanks that affect the back
pressure of pressure reliving devices.
Acknowledgements
Financial support from Neste Research Foundation
to (J-PP) is gratefully acknowledged.
References
American Petroleum Institute, (June 1995), API Standard 526, 4th
ed., API, Washington, DC.
American Petroleum Institute, (March 1993), API Recommended
Practice 520, Part I-Sizing and Selection, 6th ed., API, Washing-
ton, DC.
Heidemann, R. A., & Khalil, A. M. (1980). AICHE J., 26, 769779.
Leung, J. C. (1996). Chem. Eng. Progress, 92(No. 12), 2850.
Leung, J. C., & Nazario, F. N. (1990). J. Loss. Pre6. Process, 3,
253260.
Nazario, F. N., & Leung, J. C. (1992). J. Loss. Pre6. Process, 5,
263269.
Peng, D.-Y., & Robinson, D. B. (1976). Ind. Eng. Chem. Fundam., 15,
5964.
Pokki, J. -P., Aittamaa, J., Keskinen, K. I., Hurme, M., (1999),
Computers and Chemical Engineering Supplement, 23, 399402.
Pokki, J.-P., Aittamaa, J., & Hurme, M. (2000). Computer-Aided
Chemical Engineering-8 (pp. 811816). Elsevier.
Fig. 9. Simulated mass ow from buffer in rating mode.
buffer tank lowers the peak pressure and shifts it to
later time. Decreasing the size of the SV decreases the
ow but increases the peak pressure and shifts the peak
pressure to earlier time. These topics can be studied
only with the aid of dynamic simulation.
.