Vous êtes sur la page 1sur 5

Accurate prediction of tower relief

A comparison of conventional and dynamic simulation methods applied to a


project to debottleneck a deisobutaniser
HARRY Z HA, ABDULLA HARJI and JONATHAN WEBBER
Fluor Canada Ltd

efineries and upgraders are


often required to increase
throughput
beyond
their
initial nameplate capacity. Existing
bottlenecks must be identified and
mitigated before the unit can
process more feed. With increased
throughput, relief loads are recalculated
to
ensure
that
the
corresponding process units are
protected.
When
conventional
calculation methods are used, cases
arise where the newly calculated
relief loads exceed the existing
capacity of pressure safety valves
(PSV) on columns, or the capacity
of the existing flare system. For a
debottlenecking project, adding
PSVs or modifying an existing flare
system can be costly and impractical due to constraints on available
downtime. Conventional methods
for calculating relief loads are
generally conservative1 and lead to
over-sizing of the relief and flare
system.
Dynamic
simulation
provides an alternative method to
better define the relief system and
improve understanding of what
happens during relief.2
Dynamic simulation is rigorous
and can model many processes,
contributing to a reduction in relief
loads that the steady state-based
unbalanced heat method cannot.
The dynamic model simulates the
fluid inventory hold-up and
predicts the time before relief pressure is reached. This time before
relief can be used to take credit for
operator intervention to prevent/
mitigate the relief load. Unlike the
steady state based models, which
assume an unlimited amount of
light components being relieved,
the dynamic model accounts for

www.digitalrefining.com/article/1000961

Non-condensable
to flare

Condensers

PC

Flare

Accumulator
LC
FC

PX
72

De-iC4
Feed-1
from
alkylation
reactors
Feed-2
from OSBL

Reflux/
Product
pump

64

FC

PI

FC
Isobutane
product

HH

33

FC
1

TC

MP steam

LC
Re-boilers

Condensate

FC
De-iC4
bottoms to
debutaniser

Figure 1 System diagram of a deisobutaniser column

the depletion of light components


and estimates changes in heat of
vaporisation and temperature with
time, which can lead to a partial
loss of reboiler duty due to
decreased log mean temperature
difference (LMTD) or depletion of
liquid inventory in the column
sump.
Dynamic simulation can take
credit for the column overhead
condenser duty before the accumulator floods, reducing the relief load

through condensation. Typically,


credits are not given to the control
valve responses on over-pressure
protection in either conventional or
dynamic calculations. In cases to
consider favourable responses of
control valves to reduce relief
loads, dynamic simulation has the
advantage
of
simulating
the
response over the time of incident,
to ensure that any credits taken are
appropriate within that time frame.
Dynamic models also simulate

PTQ Q2 2014 1

Non-condensable
to flare

Condensers

PC

Flare

Accumulator
LC
FC

PX
72

De-iC4
Feed-1
from
alkylation
reactors

Reflux/
Product
pump

64

FC

Feed-2
from OSBL

PI

FC
Isobutane
product

HH

33

FC
1

TC

MP steam
Condensate

Re-boilers

Flash zone

LC

Sump
FC
De-iC4
bottoms to
debutaniser

Figure 2 Dynamic simulation configuration of the deisobutaniser column

interactions among the processing


units and provide a more realistic
overall picture of the response in
various relief scenarios. Dynamic
simulation can also help identify
potential design modifications to
reduce the relief load potential
which
conventional
methods
cannot address.
In this study, relief loads for an
existing deisobutaniser column
have been calculated using both the
conventional
unbalanced
heat
method and dynamic simulations
under various relief scenarios. As a
result of debottlenecking, the feed
to the column was increased by
18%. Calculated relief loads from
both methods are compared to each
other and against the existing PSV
capacity.
The
advantages
of
dynamic
simulations
are
pronounced when an existing
column relief system is being
rechecked for debottlenecking.

2 PTQ Q2 2014

Dynamic simulation model set-up

The deisobutaniser is one of several


columns in the refinery alkylation
unit that remove the light ends
from alkylates, to enhance the
octane number of gasoline. As
Figure 1 shows, the deisobutaniser
has 72 trays; two feed streams enter
the column at trays 64 and 33. The
column is equipped with two
baffled, once-through thermosiphon
reboilers at the bottom and
air-cooled condensers on the overhead. The heating medium for the
reboilers is medium pressure steam
(150 psig).
The bottom tray temperature is
cascaded to the flow controller on
the condensate from the reboilers.
The column sump level is
controlled by the alkylate product
flow from the column bottom. The
column pressure rides on a pushpull pressure controller on the
accumulator. The accumulator level

is controlled by the overhead liquid


product flow.
A dynamic simulation in Hysys
was set up based on the configuration shown in Figure 1 and on the
latest information on existing equipment, piping, and instrumentation
(datasheets, P&IDs, mechanical
drawings, and isometric drawings
of piping). The dimensions of equipment and piping were input into the
simulation with the exact elevations.
Control valves were simulated with
reported volume flows and openings from the datasheet. PID
controllers were used to compute
the controller actions. Relief valves
are installed on the overhead line
and a spreadsheet function was
used to calculate the required relief
area as a function of time. The PSV
sizing was based on the API
Standard.2 In particular, baffled
thermosiphon
reboilers
were
installed with an imaginary column
flash zone and a sump to achieve
the designed boil-up ratio and circulation (see Figure 2). The details of
this configuration were important in
calculating the detailed hydraulics
between the column and the thermosiphon reboilers. This allowed
simulation of the circulation and
boil-up changes resulting from the
loss of static head during column
relieving. Condensate levels in the
reboilers were not simulated and the
reboilers were assumed not to be
flooded, allowing for a conservative
reboiler duty for relief load
estimation.
The dynamic model was then
tuned to match the performance of
steady state simulation for operational conditions (P/T), feed and
product rate and quality, and the
duties of reboilers and condensers.
In the evaluation of the load from
an individual relief device, no
favourable response was allowed
by any control instrument that
would reduce the relief load
according to API standard guidelines. If the normal controller
response could act to reduce the
relief load, then the control valve is
assumed to remain in its last position before the upset.
Based on relief analysis using
conventional methods, two relieving
scenarios were identified as the

www.digitalrefining.com/article/1000961

failure)
Feed from the alkylation reactors
stops
Feed from OSBL (mixed isobutanes) stops
Overhead product/reflux pump
stops
Column bottom product continues on level and pressure as LC
fails in position
Medium pressure steam to
reboiler, flow controls fail in position (no credit taken)
25% of the normal duty is
assumed available for air cooling
condensers based on natural draft;
however, the duty of overhead
condensers is considered to be zero
once the accumulator is flooded
PSV back pressure is assumed
40% of the PSV set pressure as per
preliminary hydraulic calculation of
the flare header
PSV set pressure is 140 psig and
110% overpressure is used for PSV
sizing
Liquid level in the accumulator is
at 70 vol% before the incident
(normal operation).

Reflux failure (reflux pumps fail)


Feed from alkylation reactors
continues as normal
Feed from OSBL (mixed isobutanes) continues as normal
Overhead product/reflux pump
stops
Column bottom product continues on level and pressure as LC
fails in position
Medium pressure steam to
reboiler, flow controls fail in position (no credit taken)
Air cooler fans are assumed available until the accumulator is flooded
(then the condenser will be lost)
PSV back pressure is assumed
40% of the PSV set pressure
PSV set pressure is 140 psig and
110% overpressure is used for PSV
sizing
Liquid level in the accumulator is
at 70 vol% before the incident
(normal operation).

www.digitalrefining.com/article/1000961

900

35

UBH load
PSV Size-Dyn
Relief load-Dyn

800
700

30

600

25

500

20

400

15

300

10

200

100
0

10

15

20

25

30

PSV size required, in2

Total power failure (site-wide power

40

1000

Relief load, 1000 lb/h

potential governing relief cases for


the deisobutaniser: total power failure (TPF) and reflux failure (RF).
The assumptions for each scenario
are summarised as follows:

Time, minutes
Figure 3 Total power failure relief loads by the unbalanced heat method and dynamic
simulation at 70% initial volume

The scenarios and actions were


set up in a scheduler in Hysys
dynamic simulation. Relief cases
were studied by running the scheduler over one hours running time
(until the relief load was reduced to
an approximately constant flow
rate). The results were recorded in
small steps (0.25 second) and
summarised in plots (discussed in
the following section).

Column relief loads

For comparison, the relief loads of


the deisobutaniser were also calculated
based
on
conventional
methods for the selected cases.
Conventional methods for relief
load calculation are largely based
on steady state simulations. In
these methods, an accumulation
term is introduced to determine the
relief load based on heat and material balance at a particular instant
during the relieving process. In this
study, the unbalanced heat (UBH)
method is referenced as the conventional method.3 In this widely used
method, a heat and material
balance is produced at relief conditions to determine whether there is
excess (unbalanced) heat. The accumulation term (mass flow) is then
determined by dividing the excess
heat by the latent heat of vaporisation of the accumulated stream at
relief conditions. The reflux or the
top tray liquids at steady state are
usually taken as the accumulated
stream to obtain the physical
properties for relief load calculations. This approach simplifies

calculation of the relief load and is


designed to give conservative
results. The relief loads calculated
by the UBH method and dynamic
simulations are discussed below for
the total power failure and reflux
failure cases.

Total power failure case

Figure 3 shows the results of using


the conventional method and
dynamic simulation as a function of
time. The conventional method is
not time-dependent and the relief
load appears as a constant relief
load over time (a horizontal line in
Figure 3), while the dynamic simulation
develops
the
transient
behaviour of the system. In conventional calculations, the latent heat of
reflux at dew point and relieving
pressure is employed and an endless
supply of vaporising components is
assumed. As expected, a conservative relief load was obtained by the
conventional UBH method with the
calculated load as high as 950 000
lb/h, which was significantly higher
than the peak relief load predicted
by dynamic simulations. According
to the dynamic simulations, relief
does not occur until 6.5 minutes
after the TPF event occurred after
the column overhead pressure
reached the PSV set pressure. The
peak relief load is reached approximately five minutes after the relief,
beyond which the relief load
decreases to below 100 000 lb/h
within 21 minutes. The calculated
relief area corresponding to the peak
relief is 30.6 in2 which requires a T

PTQ Q2 2014 3

40

PSV Size
Relief load

400

35

350

30

300

25

250

20

200

15

150

10

100

50
0

10

15

20

25

30

PSV size required, in2

Relief load, 1000 lb/h

450

Time, minutes
Figure 4 Total power failure relief loads by dynamic simulation at 55% initial volume

and a P sized API PSV, while the


relief load calculated by the UBH
method requires two T sized PSVs.
However, the existing deisobutaniser only has one 8T10 API PSV (26
on
its
overhead
line.
in2)
Considering the challenges and
substantial costs to add an additional PSV and modifying flare
system, mitigations were investigated by lowering the initial liquid
level at the overhead accumulator to
minimise the impact on the project.
The initial liquid level in the
overhead accumulator was a key
parameter for column relief calculation. In conventional methods, it is
common practice to assume the
overhead drum and condenser will
be flooded as a result of losing the
reflux and product pumps (unless
there is significant hold-up volume
allowing adequate operator intervention time in the overhead
drum). In dynamic simulation, the

pump failure scenario can be simulated and credit taken for the time
it takes the overhead drum to flood.
This time is related to the liquid
level in the drum at the beginning
of the relief scenario. The original
normal liquid in the accumulator
was 70% by volume. An initial
liquid level of 55 vol% was studied
in dynamic simulation and the
result is shown in Figure 4. With
the lowered liquid level in the
drum at the beginning of relief, the
time to flood the accumulator is
postponed
by
two
minutes
compared to the 70% initial level
case. As a result, the peak relief
load was reduced to about 430 000
lb/h. The corresponding required
relief area is 22.6 in2, and the existing T sized PSV is now sufficient.

Reflux pump failure case

Figure 5 shows the results of relief


analysis for the case of reflux/

Relief load, 1000lb/h

700

UBH load
PSV Size-Dyn
Relief load-Dyn

600

30
25

500

20

400

15

300

10

200

100

PSV size required, in2

35

800

0
0

10

15

20

25

30

Time, minutes

Figure 5 Reflux failure relief loads by unbalanced heat method and dynamic simulation
at 70% initial volume

4 PTQ Q2 2014

product pumps failure. The relief


load calculated by the UBH method
remains high (757 000 lb/h), while
the dynamic simulation predicted a
peak relief load of 562 000 lb/h 7.3
minutes after the pumps failed. The
required relief area for the relief
loads calculated by the UBH
method and dynamic simulation
are 40 in2 and 30.1 in2 respectively,
both exceeding the available area of
the existing PSV (26 in2). Compared
to the dynamic simulation of the
TPF case, the reflux pump failure
case has a similar peak relief load
but occurs three minutes later than
the TPF case. This is due to the
continuous feeds to the column and
the liquid on the trays being heated
up more slowly than in the TPF
case. (The feed temperatures are
lower than the temperature of the
liquid
inside
the
column.)
However, the relief load decays
much more slowly than in the TPF
case after the peak; the relief load
was still significant at 200 000 lb/h
even after 60 minutes. The flatter
relief load profile is attributed to
more light components available
from continuous feeds under the
reflux pump failure case.
Similar mitigations to the TPF case
were also considered for the reflux
pump failure case (see Figure 6). As
a result of reducing the initial liquid
level in the overhead accumulator
(from 70% to 55% by volume), the
column relief condition was delayed
by 2.3 minutes and the relief load
was reduced to 462 000 lb/h which
required a relief area of 24.8 in2. The
existing PSV (26 in2) will be able to
take the reduced peak relief loads.
Considering the challenges and risk
in modifying the existing relief and
flare
systems,
the
mitigation
approach (lowering the initial liquid
level in the overhead from 70%
volume to 55%) is recommended to
minimise the impact on the debottlenecking project.

Summary

Conventional methods for column


relief load calculation are typically
more conservative than in dynamic
simulation due to the conservative
assumptions inherent in the calculations. Conventional methods do
not account for the compositional

www.digitalrefining.com/article/1000961

450

Relief load, 1000 lb/h

35

PSV Size
Relief load

400

30

350

25

300

20

250
15

200
150

10

100

50
0

10

15

20

25

30

PSV size required, in2

500

Time, minutes
Figure 6 Reflux failure relief loads of dynamic simulation at 55% initial volume

changes on the trays and the


system volumes contribution to the
transient availability of cooling and
heating. They also ignore the time,
temperature and pressure dependency of heat input or heat removal
from the system. In the cases of
plant revamp or debottlenecking,
conservative relief loads from
conventional methods will likely
call for the addition of PSVs and/or
the modification of the existing
flare system, which adds substantial cost and risk to the project.
Dynamic simulation simulates the
actual composition and inventory
changes inside the column and
accumulator and provides more
accurate predictions on relief loads.
Typically, the relief loads predicted
by dynamic simulation are less
than those calculated by conventional methods. Depending on the
system, the reduction in calculated
relief loads can be significant (by
>50%). For plant revamp or debot-

www.digitalrefining.com/article/1000961

tlenecking projects, this reduction


in relief loads can result in significant savings for the project.
This case study illustrated such an
example. Employing dynamic simulation for an existing deisobutaniser
has reduced the calculated relief
loads by more than 35% for both
TPF and reflux pump failure cases,
compared to the conventional methods. This study also demonstrated
that the initial liquid level in the
overhead accumulator is a key
parameter in relief load calculation.
When the initial liquid level is set at
70% volume (as it is in current operation), the relief loads predicted by
dynamic simulation exceeded the
available capacity of the existing
PSV. Mitigation approaches were
studied by lowering the initial liquid
level in the overhead drum to 55%
volume. As a result, the relief loads
were further reduced by 26% and
18% for the TPF and reflux pump
failure cases, respectively. With the

recommended mitigation approach,


the new predicted relief loads are
within the capacity of the existing
PSV. Therefore, the risk and cost of
modifying the existing relief and
flare systems are minimised and
potentially avoided.
References
1 Chittibabu H, Valli A, Khanna V, Calculating
Column Relief Loads, PTQ, 55-65, Q2 2010.
2 API RP 520: Recommended Practice for the
Design and Installation of Pressure Relieving
Systems in Refineries, Part I (Sizing and
Selection, 2008) and Part II (Installation, 2003),
American Petroleum Institute, Washington D.C.
3 Sengupta M, Staats F Y, A new approach to
relief valve load calculations, 43rd Proceedings
of Refining Section of American Petroleum
Institute, Toronto, Canada, 1978.
Harry Z Ha is a Senior Process Engineer/
Specialist with Fluor Canada Ltd, Calgary,
Alberta, Canada. He holds a masters degree
in environmental engineering from Hong
Kong University of Science and Technology
and a PhD in chemical engineering from the
University of Alberta.
Email: Harry.Ha@Fluor.com
Abdulla N Harji is an Executive Director of
Process Technology, at Fluor Canada. He holds
a BSc degree in chemical engineering from
Loughborough University, UK.
Jonathan Webber is a Process Engineer with
Fluor Canada. He holds a PhD in process control
from Dalhousie University and a masters in
biotechnology from McGill University.

Links
More articles from: Fluor
More articles from the following
categories:
Mass Transfer & Separation
Pressure Drop Control
Process Modelling & Simulation

PTQ Q2 2014 5

Vous aimerez peut-être aussi