Professional Documents
Culture Documents
Richard Irons
July 2009
Contents
Page
1 Executive Summary 3
2 Background 4
3 Type 1 – Simple Knock Out Pot 6
4 Type 2 – Axial Cyclone 7
5 Type 3 – Conventional Cyclone 10
6 Type 4 – Bottom Discharge Cyclone and External Disentrainment 13
7 Separator Design Summary 16
8 Water Sprays 17
9 Drain System 18
10 Separator Selection Flow Chart 19
11 Appendix attachments 20
Generally moderate cleaning of the vapour to minimise vapour line and heater scaling can be
achieved at minimal expense. More improved cleaning will require the installation of a
separation device/vessel which may require more significant piping and structural
modifications. Separation, collection and routing of the contaminated liquid drain prior to the
heaters will also increase the cost and complexity of a retrofit solution. It is therefore important
for the refinery to understand their requirements and their budget before progressing to an
engineered solution, a rough decision flowchart has been include on page 18 to assist in this
process.
a. Thermal Performance
If improving the condensate quality is not important then the main reason to improve the flash
vapour quality will be to minimise scaling in the vapour lines and the heater shells, and any of
the four solutions detailed in this report would be appropriate. However the cost of each solution
progressively increases from Type 1 the cheapest, to Types 3 and 4 the most expensive, each
solution has its specific advantages and issues and these are tabulated in section 7 on page16.
The type of carryover to be removed is also an important consideration in selecting the type of
disentrainment device to be used. If the carryover has excessive quantities of sand then a
separation device with lower internal velocity (like a traditional cyclone design) may be required
to prevent excessive erosion.
b. External disentrainment
Most flash vessels are equipped with an internal disentrainment device. Typically older vessels
have a spiral vapour exit where the flash vapour exits the flash chamber through a small
doorway (to accelerate the vapour) into a spiral annulus where the intent is for carryover
droplets to impact on the outer wall. The carryover drains across the floor and back down into
the flash chamber. Later type vessels have a central cone discharge from the flash chamber to
accelerate the vapour with a “mushroom” plate above to impact the carryover droplets. Typically
the “mushroom” plate will have a serrated edge to collect the droplets into streams to minimise
re-entrainment. Again the carryover drains back into the flash chamber.
Both these types of internal disentrainment can scale heavily particularly as unit throughputs
have increased, increasing the volume of carryover liquor and slurry. Typically flash vessels
are taken offline every 12 to 18 months for statutory inspections at which time the scale is
removed. This may take anything from a few days to several weeks during which time the
vessel is bypassed and this generally represents an energy loss.
Design Parameters
18 TPH
Predicted dP = 0.4 kPa
Predicted capture efficiency;
10μm bauxite (wet walls) = 90%
10μm droplets (wet walls) = 65%
Measured capture efficiency;
Particles = 90%
Droplets = 70%
The model shows the particles entering with a linear velocity into the spiral accelerator and
being spun to the outside of the pipe, the typical velocity in the spiral section is around 45m/s.
The particles remain on the wall as the pipe section is increased. The increased diameter
reduces the particle velocity in the outer annulus and the floor slope ensures that the carryover
particles and excess water droplets report to the side drain. The cleaned vapour exits the
separator through a centre pipe which is the same diameter as the inlet pipe.
The high velocity in the spiral section can be an issue if there is too many sand particles in the
carryover as this can lead to an excessive wear rate. Holing out the spiral section during normal
operation would result in the vessel having to come off line to affect a repair resulting in a
production loss. The photograph on the following page shows the installation at Alumar.
The performance of the Alumar cyclone is detailed in the report written by Carlos Cahete &
Steve van Lierop, the report is attached in the Appendix on page 20. In summary the first stage
CRD condensate conductivity dropped from 275S/cm to 81S/cm better than the design target
of 130S/cm. The cyclone reduced the overall CRD condensate conductivity from 250S/cm
to an average of about 184S/cm.
Even though the axial cyclone experience at Alumar was promising there were concerns on
using this technology in the WA refineries due to the sand content of the vapour and the higher
vapour line velocities.
The diagram below shows the CFD modelling of the axial separator design for the 400NB (16”)
vapour outlet line from the Pinjarra 11# flash tank. Note that the predicted performance is better
than that for the Alumar design – because of the higher velocities but the additional
performance comes at the cost of increased pressure drop.
Design Parameters
22.9 TPH
Predicted dP = 1.5 kPa
Predicted capture efficiency;
10μm bauxite (wet walls) = 95%
10μm droplets (wet walls) = 81%
Comments:
dP and separation very good due to the
high internal velocity
Design would only just fit into existing
piping
Inlet velocity is 68 m/s! Erosion through
vane set likely to be very high.
Sao Luis trial indicated that an inlet
velocity closer to 30 m/s would still give
acceptable performance & would give
lower dP and erosion
Cyclone could be redesigned to use
larger feed pipe (e.g. 600 NB), but
would struggle to fit within the space
constraints at PJ.
At Pinjarra the higher internal velocities generated in the axial cyclone was likely to give an
unacceptably high wear rate in the vessel body leading to potential loss of containment. Any
loss of containment from holed a separator would be unacceptable from an OH+S perspective
and also from the energy and production losses incurred while repairs were affected – this
would have to be done with the flash tank out-of-service. The vapour separators are designed
without separate isolation or bypass valves to minimise valving/isolation issues and
maintenance costs. The high pressure drop was also unacceptable as this was additional to
the pressure loss across the existing internal disentrainment system.
To minimise the internal velocities, and therefore reduce any wear issues, and to minimise the
overall pressure drop a conventional cyclone was designed. CFD modelling was used to
optimise the design. The modelling results are shown on the following page.
The cyclone was installed on the 11# vapour line to the mill heaters such that if there were
issues with reliability or wear the cyclone could be isolated without impacting on the flash tank
availability.
The cyclone averaged a dp of 0.85kPa at 18T/h of vapour and 1.0kPa at 26T/h (inlet velocity
of 20m/s) – better than predicted by the CFD modelling.
The cleaned vapour from the cyclone was subsequently mixed with vapour from other units so
it was not possible to sample condensate from the cleaned stream. Isokinetic sampling from
the cyclone inlet was also not possible due to the design of the pipework and difficulty of
sampling a stream with very small carryover content. This meant fully assessing the cyclone
capture efficiency was not possible. However as the spray flow was metered samples were
taken of the cyclone drain/underflow. The average conductivity of the underflow was 990S/cm
indicating a good capture of carryover liquor droplets.
There was an anecdotal report that the vapour valves on the clean vapour side could still be
operated some time after a maintenance overhaul where previously these valves were made
in operable due to scaling after only a short time.
To manage the underflow drain the cyclone was fitted with two PLC controlled butterfly valves
and a water quench system. The double valving enabled the cyclone to be drained of collected
liquid/sand without passing steam. Similar to the Alumar experience the drain system was
problematic. There was some flashing across the valves which damaged the valve seats
leading to excessive wear of the seat and downstream piping. Damage to the shaft seal area
The design was fine-tuned using CFD modelling to maximise the capture efficiency and
minimise the pressure drop.
The first new BOT was successfully commissioned on 9th January 2009. The combined CRD
flash condensate conductivity was around 120 to 140μS/cm with the liquor flow to the unit @
1000kL/h LDCH and all flash tanks online, acceptable but higher than expected. Further
investigation found that the high conductivity was due to carryover from the secondary flash
tank which was operating overlife (scaled), and a tube leak from an upstream heater.
The unit was put into a flash tank bypass on the 14th January and the CRD flash condensate
conductivity improved to around 75μS/cm once the secondary flash tank was out-of-circuit. This
figure was better than expected indicating that the disentrainment cyclone was working well. A
reconciliation of the cyclone drain and condensate conductivities indicated the CRD
With the system operating with all flash tanks online underflow samples were taken from the
cyclone to assess the captured material. Typically the conductivity of the underflow was around
12μS/cm and the solids content around 0.7gpl for a spray flow of 1kL/h. This indicates that
during normal operation the carryover from the new upflow flash tank is minimal and that the
cyclones are only required during flash tank bypass duty.
A sufficiently large solids sample was collected to analyse the size distribution. The data shows
that the cyclone is capable of capturing particles of <10μm which is consistent with the CFD
design.
Water Spray and/or Conventional cyclone Axial Cyclone Downflow cyclone BOT external chamber
Knock out Pot
For Cheapest option - High separation efficiency. Good separation efficiency High separation Low pressure drop (when
can be installed at Proven technology. with low pressure drop. efficiency. clean).
minimal cost and in Moderate wear so internal Proven technology. Proven technology. Low wear.
short time period. wear plates required. Can be good fit into existing Moderate wear so Water spray installation
Minor modifications Moderate “surge capacity” – pipework if sufficient internal wear plates possible but positioning more
to existing volume = 8m3. elevation difference between required. difficult than other options
equipment required. Can improve CRD flash tank and CRD system. Relatively easy fit into due space limitation between
Can possibly condensate quality if drain Can improve CRD existing CRD vapour vessels.
achieve 50% of the is re-routed. condensate quality if drain is piping. Good “surge capacity” -
scale reduction of Could possibly be designed re-routed. Moderate “surge volume = 11 m3.
the other separation to fully replace internal capacity” – volume = Descale can be remotely
systems. disentrainment but not yet 9m3. done but still not easy.
Can improve CRD proven for this. Can improve CRD
condensate quality Can be easily remotely condensate quality if
if drain is re-routed descaled. drain is re-routed.
but this can Avoids height limitations of Can be designed to
increase cost. axial cyclone design but top fully replace internal
Requires no major discharge means this disentrainment.
re-work of structural design is taller than bottom Can be easily
supports or discharge cyclone option. remotely descaled.
foundations. Avoids height
limitations of axial
cyclone design.
Against No advantage to Complex geometry and Potentially high erosion for in Requires external A departure from proven
flash tank TAT. pipework can make retro- applications with excessive support structure so technology.
No improvement to fitting in brown field sand carryover like WA. can be expensive. Poor separation efficiency
OH+S issues situations more difficult. Longer length to diameter Large crane required compared to cyclone and
associated with Requires external support ratio can make installation in if remote descaling is axial separators.
descale. structure so can be some sites difficult due to required. Support structure can be a
expensive. height constraints between problem due to high weight
Higher pressure drop than the flash tank vapour outlets when scaled especially if
axial and downflow designs. and the CRD heater inlet. mounted on flash tank roof.
Large crane required if Limited “surge capacity” - Large crane required if
remote descaling is volume = 3 m3 for 600 NB remote descaling is required.
required. inlet design
8. Water Sprays
In some cases much of the scale reduction benefit of installing external disentrainment can be
achieved by installing just the water spray, particularly if the vapour piping is convoluted
providing a lot of impingement surfaces.
For optimum performance it is important the correct type of spray nozzle is used. The spray
nozzle should be a “full cone” type with a 90º included spray angle. The spray should be
installed such that the nozzle is in the centre of the pipe spraying axially down the pipe. To
facilitate easy removal and replacement the spray nozzle should be screwed to the condensate
supply pipe which should be flange mounted to the vapour piping. This allows for easy spray
assembly removal even if the nozzle assembly is covered in scale. A typical installation
(Pinjarra BOT) is shown on detail 1 on drawing PJ076036 in the appendix. The drawing shows
the spray nozzle fitted to a 20mmNB XS pipe. The piping inside the vapour line is XS to ensure
the pipe doesn’t bend with the vapour velocity. In large vapour lines it may be necessary to
provide additional support to the pipe by bracing it with some flat bar.
In applications where the required spray flow is high and there is a limitation on the water supply
pressure the angle type full cone spray may have insufficient capacity, in these instances a
hollow cone type spray can be used.
The nozzle used in the Pinjarra BOT application is a Spraying Systems™ ½” GGA –SS 40.
This is a ½” full cone angle type nozzle, the spray angle at the nominal flow is around 90º over
a wide range of pressures, it has a large clear orifice so as not to be susceptible to blocking
with any sand present in the condensate supply.
Selection of the correct spray nozzle is critical, the photo below shows a “pig- tail” type spray
trialled in a digestion flash vapour application. This nozzle was in operation for 2 months and is
fully scaled. The main issue with this type of spray nozzle is the high surface area that is
presented to the “dirty” vapour during normal operation, this leads to a progressive build up of
scale which eventually restricts the water flow.
The photo below shows the recommended type of spray nozzle, in this case a Spraying
Systems™ ½” GGA type. It is more compact than the “pig tail” type which means it can be
inserted into the centre of a pipe to spray in the direction of flow through a smaller flange making
the installation and maintenance cheaper and easier.
9. Drain Systems
Careful consideration must be given to the drain system for the removed carry over and excess
spray water. As the drain contains carried over green liquor and the temperature is cooling the
drain systems are susceptible to scaling.
The first important criteria to be considered is where will the drain liquor report to. If the drain
liquor returns to the liquor circuit this can simplify the system but will add dilution. This is the
system that was adopted by Pinjarra for the blow off tank disentrainment cyclones, where the
cyclones drain directly back to the blow off tanks with no in-line valving. The drain returns to a
point below the slurry level to ensure no vapour is vented up the drain line to the cyclone.
However the in the Pinjarra installation the tank and drain can be isolated for cleaning if required
without a production impact due to the spare blow off tank. The absence of inline valves makes
the system robust.
In the Pinjarra top discharge cyclone, and the Alumar axial cyclone, the separator drains do not
return to the liquor circuit, both systems discharge the collected carryover and condensate to a
waste water system. In both these installations drain valve reliability is an issue. Careful
attention to this valving is required during detail design to ensure the system is robust and can
be maintained without having to take the disentrainment device offline.
Axial Ciclone.doc
Top Discharge Cyclone as installed on the Pinjarra Unit 4 Secondary Flash Tank
PJ060989.pdf
PJ076032.pdf
PJ076306
Condensate Supply Typical Riser BOT arrangement R3.pdf