PETROCHEMICAL DEVELOPMENTS
be a eliae
Distillation column relief loads—
Part 1
Compare conventional calculation methodology with dynamic simulation
P. L. NEZAMI, Jacobs Engineering, Houston, Texas
‘most complex relief load calculation. The dynamic nature
of distillation columns and compositional changes along
the column height make it very difficule to accurately establish
relief loads for various contingencies. Additionally, the inability
of regular (steady-state) process simulation software to predict
column behavior in a nonsteady-state condition forces design
engineers to create analytical methods that by nature are conserva-
sive. This could result not only in oversized relief valves but also
in unnecessarily large and costly flare systems, which in turn can
jeopardize project viability
‘The enormous differences between distillation systems, such
as column controls, types of condensers and reboilers, heating
media, pumparounds and side reboilers, ete., make
ic impossible to create a universal method for all distillation
columns. The suggested techniques are at best a series of general
guidelines and criteria, set forth to provide directions to evaluate
each individual ease
E stimating the relief rates for distillation columns is by far the
Common methods. The two most prevalent shortcut tech-
niques, which may or may not be conservative and generally
used to estimate order of magnitude relief rates, are:
Flash drum approach. This could be used to estimate the
relief rate in a loss of cooling/condensing scenario. In this
method, the feed stream is flashed at relieving pressure with
additional heat input equal to the reboiler duty.
Gross overhead vapor. This is usually used to establish a
basis for the flare and flare header design at early stages in the
projects, and it is the simplest way to roughly estimate relief
loads. Although this method seems to be conservative for most
cases, it has been argued that it could result in undersized relief
valves.!
For better results, one must analyze relief scenarios in detail
‘on a case-by-case basis. The most comprehensive conventional
method is to estimate relief loads based on mass and energy
imbalance (accumulation) in an upset condition. The bases and
assumptions for this method are:
1. At elieving conditions, feeds, products and reflux con
positions, as well as top-tray liquid and bottoms compositions,
are unchanged.
column trays are at vapor/liquid equilibrium at relief
pressure.
3. Except for the feeds, all streams leaving and entering the
column are at vapor/liquid equilibrium at relieving pressure.
4, Vapors may not be accumulated in the column (after the
column reaches the relief pressure) and must leave the system
via a relief valve. Liquids could accumulate in the system by the
rise or fall of liquid levels.
5. Liquids can absorb heat whether they leave or stay in
the system.
6. Credit may be taken for product sensible heat absorption
when the feed enters the column below its bubble point at the
relieving pressure.
7. The energy imbalance resulting from an upset is converted
to mass (vapor) using top-tray liquid latent heat of vaporization,
8, The vapor distillate control valve, if applicable, stays at
its position. Credit may be taken for vapor distillate unless its
path is blocked.
9. The vapor portions of the feed streams, flashed adiabati-
cally at relief pressure, directly contributes to the relief rate.
10. Credit may be taken for reboiler temperature pinch, if
light materials do not reach the column bottom.
11. Any safety margin used in the actual design must be
considered in the relief rate calculations.
12. The properties of the vaporized top-tray liquid at bubble
point and relief pressure are used to size the relief valve,
‘The relief rate can be defined as:
Wa Wat We- Wy Wo- Wi ay
where:
Relief rate
Wp Reboiler and side reboiler load contributions
‘ced vapor phase contributions
= Vapor di
ndenser and pumparound credits
juids enthalpy imbalance.
llate credit
Wa=
Dynamic simulation. Nevertheless, the best available tech-
nique for distillation column relief rate calculation is dynamic
simulation. Dynamic models use a set of mass and energy con:
servation equations that account for changes occurring over time.
Unlike steady-state simulation, these equations include an addi-
tional accumulation term that is differentiated with respect to
time. The accumulation rate of mass is:
‘Mass flow into system — Mass flow out of system )
And the accumulation rate of total energy is:
Flow of toral energy into system — Flow of otal energy out
ofsystem + Heat added to system across the boundary + Heat
generated by reaction ~ Work done by system on surroundings (3)
HOROCARBON ROCESING APR 2008 | 115PETROCHEMICAL DEVELOPMENTS
the mass and energy
conservation equations allows the dynamic model to rigorously
calculate composition changes at each stage and to modify vapor!
liquid equilibrium over time. It also allows integrations of column
flowrates, pressures and temperatures with respect to time. The
results lead to a much more accurate relief rate
Case study. A comparison between the results of the con-
ntional methed with those of the dynamic simulation for a
typical distillation column in a loss of condenser scenario will
be presented. Fig. 1 shows the flow diagram of the distillation
column system under evalu
Ina loss of condenser scenario, the reflux drum liquid level
drops causing the level control valve to close, Reflux continues at
constant rate until ehe drum runs dry. The feed continues at a
constant rate since its pressure, upstream of the control valve, is
hhigher than the relief pressure, With no liquid distillate product,
the only place for light macerials in the feed is the colurnn bot-
toms; hence, no credit could be taken for reboiler temperature
pinch. On flow contro, the steam control valve opens wide and
the reboiler chest pressure equalizes with the steam header pres-
sure (a zero pressure drop is assumed across the control valve).
Clean reboiler duty is used for both calculations.
Conventional method. Table 1 is the summary of the
stream properties and column parameters at normal and upset
conditions. Stream properties in upset conditions were obtained
normal stream compositions:
+ Feed:
+ Vapor distillate
* Decant water—bubble-
Wy = 403,980 lb/hr
W,=0 lb/hr
vapor a relieving pressure
Wy= 9,128 Ib/he
We= 0 lb/hr
Wi, = 179,992 Ib/he
The relief rate would be the sum of the above calculated
values:
W= 214,860 Ib/hr
Dynamic simulation. A dynamic model was developed for
the column, Fig. 2 shows the simulation flow diagram for it
The following parameters were set to generate the initial values
of dynamic simulation variables:
* The reboiler UA was set to clean UA value fora zero fouling
factor on both sides.
* The reboiler chest pressure was adjusted to obtain normal
reboiler duty. This emulates pressure drop across the steam control
valve at steady-state conditions.
* The reflux drum and column bottom sump dimensions
were set co actual values co simulate liquid level variations in a
dynamic mode.
+The condenser was set ro constant medium temperature mode.
The initial setting of dynamic parameters, such as reflux
drum and column bottom sump dimensions, reboiler heating
‘TABLE 1. Stream properties and column parameters
by performing the following flash calculations on the basis of Normal Upset
3136 sai
adlbai ash at rele pesure as aa
+ Laqullidielare Bubbles ple lash at qeliet pesture chalets oe ae
flash at relief pressure (Reteioes proaict Noma Uys
* Bottoms product—bubble-point flash at relief pressure. __‘lowrate, br 262376 252,376
Based on the calculation results listed in Table 1 Temperature, °F 328.7 393.3
Enthapyotul ore 583
‘por dtiate Normal Upset
Fowate ahr 750 ae
The vapor distillate pressure control valve at its normal posi- _Femperstre, °F 1174 1705
tion (normal Cy) can pass 9,128 Ib/hr of the column overhead Enthalpy, Btu =1%.9 at
gu istate Normal Upset
Fowate tir 120877 7
can ita TT
nay Bt Sarr “16
Feamulaton ate ir a 15.798
Decant water Normal Upset
oma ir 7 7
Tenpertue °F iin 8,
nth Btu “ans “Eee
Feamulaton ate he a 3B
Column parameters Normal Upset
ondeser yon Bae “S502 3
Reboler dy Btu 398 336
iss inbalnce ir 0 T5872
Enhalpy alae, lion Bahr 006 21377
Top igi ater eat, tb T24
Reif rat, he 214860
116 | APR 2008 ivoROcARBON PROCESSINGPETROCHEMICAL DEVELOPMENTS
medium temperature, column tray hydraulics, ete., are part of
the required data in steady-state simulation. The parameter
values will be carried over to the dynamic simulation as initial
variable values when the steady-state simulation is exported to
the dynamic model
In the dynamic model, the condenser failure was initiated
after 10 min, of normal (steady-state) run, At the same time, the
medium temperature was set to the steam header
temperature to simulate a zero pressure drop across the steam con
ive. The simulation was run for ewo hours, and the results,
aptured and recorded (Figs. 3-8).
re some highlights of the simi
overhead pressure starts to rs
introduced. It peaks at 165 psig 12 min, later.
+ Liquid distillate stops 5 min. after the condenser fails (this
interval depends on the level controller parameters: proportional
gain, integral time, derivative time, process variable and output
ranges). The 5-min. duration, in this example, is based on the
simulation’s PID controller default parameters,
tion results:
after the upset was
Psvi rot
‘Max allowable accumulation pressure is
asad on 110% of design pressure
— Max allowable accumulation pressure, 165 psig |
3
s
(Overhead pressure, psig
60
Time,
Continued
Choose ABB as your
Partner for Refinery Proc
FTIR Solutions
FTIR Analysis Solutions Addressing the
Needs of the Modern Complex Refinery
line & Diesel Blending
@ Hydrou
Downstream Petrochemicals
The Best Servic
Full Analyser Engineering and System Integration
B startup, Commissioning and Lifetime Support
® Analytical Method Development and Calibration
Modelling Services
Bom
Audit Services
‘Support, Validation and Performance
Select 186 at www.HydrocarbonProcessing.com/RSa Saeame PETROCHEMICAL DEVELOPMENTS
40
0
$00
320
S20} —}
#150] ii istiate
3 100 |
eve
a
si
°
Vapor istiate
3 80 30 ¥9
Time, min
Simulated stream flowrates
Teboler chest temp.
Reboil
Bottoms tmp.
‘Overhead temp,
0 30 60 90 12
Time, min,
Se
5 Bottom sump
Reflux drum
Cr)
ed
Reboiler duty
Time, min
SEI
3 8 0 720
Time,
FIGI!) Column relief rates for simulation
+ Reflux continues for another 25 min. at normal flow rate
then practically stops. This is when the rel
«The relief valve starts relieving 5 min. after the upset, exactly
when liquid distillate completely stops.
* The relief rate peaks at about 161,200 Ib/hr 12 min. after
che upset and keeps at almost a steady rate for 16 min... Then
it drops for about 7 min. to 8 min. and reaches a new steady
118 | APRIL2008 HYDROCARBON PROCESS
rate of about 122,000 Ib/hr very close to the normal liquid
distillate rate,
* The overhead temperature starts to increase at upset and,
after about 6 min, to 8 min., it reaches a temporary stable con-
dition. It stays steady so long as the relief rate is at its peak. As
the relief rate drops, the overhead tempera
its new maximum and stays constant through the rest of the
simulation time
* The vapor distillate rate incteases to double the normal
flowrate immediately after the upset. Its mass flowrate keeps
rising slowly due to the change of overhead molecular weight
and reaches its maximum value at the end of the simulation
time. The vapor distillate rate is very comparable to the one
calculated in the conventional method.
* Reboiler duty spikes to its maximum, right when the upse
is initiated. Ie drops immediately and keeps flat through the 25
min. of maximum relief rate; then it drops again, this time, to
its minimum steady level.
+ The column reaches a new steady-state condition approxi-
mately 30 min. after the upset with a constant relief rate of
122,000 Ib/hr.
he simulation results are based on hysteresis (opening and
losing curves) of atypical relief valve in compressible fluid ser-
vice. The relief valve hysteresis is presented in Fig. 9Pressure, psig
The major advantay
venti shod, is the accuracy of calculated relief rates. The
ability of the dynamic model to integrate simulation variables
of dynamic simulation, over the con:
jous column parameters
would result in much more realistic relief load values,
Another advantage is engineering man-hours and project
schedule time savings. Given that the conventional method
requires a separate detailed analysis for each relief case, dynamic
simulation can save a great deal of engineering time, since a
SPECIALREPORT
single dynamic model can be used forall various relief scenarios.
The third advantage of a dynamic model is that, for a great
majority of the cases, the relief loads calculated by dynamic
simulation are smaller than the ones calculated by using the
conventional method. This could result in a considerable cost
saving, particularly in cases where there are limitations in the
flare system capacities.
In the example, the difference is greater than 33%, which
could even be higher (up to 41%) if, instead of a single large
relief valve, two smaller ones were used. With multiple relief
valves, the allowable accumulation pressure would increase
to 116% of the design pressure and the combined relief loads
would peak at 152,600 lb/hr. At the end it is worth mentioning
that, if the conventional method relief loads are smaller than the
ones calculated by dynamic simulation, one can be certain that
he conventional method results would lead to an undersized
relief valve. 4P
LITERATURE CITED
Bradford, M. and D. G, Durrett
ee eee
ea Mee
OER eee ee We re
oe Oy Ren kama a
Eee gC Mra Reco er tlr
er re
Plant Locator, a geographical listing of HPI
plants; the Engineer/Constructor Directory
Cre kemeontc tis amt
TRG a
S Oise RU
STE eau cea)
a
Pest
eae Ana Crs
eect iee sone
Cas sact
ey
HYDROCARBON
PROCESSING