Targeted liquid injection from a secondary pumparound (PA)
effectively compensated for a shed-deck vapor distribution
problem in a fluidized-bed coker. This case study shows how the
installation of an unique piping and lance design achieved
targeted liquid injection, quenched vapor hot zones and reduced
resid entrainment into the heavy coker gasoil (HKGO)
The Valero Benicia refinery, owned by Valero Refining CompanyCalifornia,
operates a 28,000-bpd fluidized-bed coker.a The
coker-reactor-effluent vapor passes through a single cyclone
discharging into a scrubber that contains several rows of
multi-pass shed decks. The scrubber-bottom liquid is pumped
both as a recycle stream to the reactor and as PA liquid to the
top of the sheds that feed through a slotted-ladder-type pipe
distributor. The top section of the scrubber contains
structured-grid packing washed with HKGO. After passing through
the reactor and scrubber, the HKGO and light gasoil (LGO)
products are separated by the fractionator, as shown in
1. Scrubber and fractionator PFD.
In a coker, the scrubber quenches the thermally
cracked-reactor products and condenses the heaviest product
fractions for recycling back to the reactor. The void space
between the cyclone outlet and the bottom shed row acts as a
spray chamber with good liquid distribution. It can provide a
significant part of the PAs heat transfer and solids
removal. The remaining heat transfer and solids scrubbing
occurs in the shed decks.
While coking and fouling of the scrubber occurs over a
typical cycle, at the midpoint of the run, HKGO product quality
should be maintainable without undercutting. In this case
history, halfway into the cycle, significant residue carryover
into the HKGO was observed, along with accelerated increases in
the scrubber-grid pressure drop. Fig. 2
illustrates the steep slope and rapid degradation of the HKGO
product quality from resid contamination. The impact on
downstream units catalyst life and conversion would have
required an early turnaround for the coker. A troubleshooting
effort focused on identifying and resolving the entrainment
2. Scrubber grid DP growth over time
as compared to previous run.
ROOT CAUSE FOR RESID CARRYOVER
Resid is routinely entrained from the scrubber PA section.
The resid carryover into the fractionator is typically measured
at less than 5 vol% of the HKGO. As the HKGO becomes more
contaminated with resid, downstream GO-filter change outs
increase. Additionally, the higher resid percentages can
negatively impact runs on the downstream cat-feed hydrotreater
and fluid catalytic cracking (FCC) units. The elevated resid
content level of the HKGO carries both operational and economic
The scrubbers structured packing acts as a guard
against resid carryover. The grid provides a surface for
de-entrainment of liquid carried up from the scrubber PA; it
also enables heat transfer between HKGO wash and the scrubber
vapor. Efficiency loss of the grid packing can be a function of
either vapor maldistribution into the grid or liquid
maldistribution over the grid.
Structured packing is very sensitive to liquid
maldistribution. Skillful designers pay careful attention to
liquid distributor details. Liquid maldistribution of the HKGO
wash can result in poor contact of the vapor and liquid through
the grid and failure of the scrubber grid to properly wash the
Structured packing is also very sensitive to vapor
distribution. In this coker scrubber, high localized
C-factors (Cs) created channels through the
structured packing and entrained liquid from the top of the
bed. Both potential root causes of resid entrainment were
The structured packing is washed by HKGO through a spray
header.b To meet the unique anti-fouling
requirements of the coker, the spray header uses three separate
laterals, each with its own column nozzle. Each lateral has
several large orifice spray nozzles. There are no 90°
elbows on the laterals to allow for external flexible-lancing
should they become plugged.
Operating data review
While the overall slope of the scrubber grid pressure drop
was trending higher during this cycle than in the previous
cycle, a distinct step change was observed at the approximate
time that the resid content in the HKGO increased (Fig.
3). An event with the loss of HKGO flow to the grid
spray headers could result in coke formation within the bed.
Operating data were evaluated to determine whether any
instances of a loss in wash-oil rate preceded the higher resid
entrainment or step change in the scrubber-grid pressure drop.
There were no obvious losses in wash oil that may have created
a dry-out situation.
3. Scrubber grid DP and resid in
HKGO step change occurring at the end
of August 2012.
Failure of a spray nozzle or full lateral can leave areas of
the grid bed unwetted, resulting in localized dry, hot zones
that are subject to coking. Plugged or damaged nozzles would be
evidenced by a deviation in pressure drop from design. A
pressure survey was conducted to determine the actual nozzle
pressure drop. Comparison of field data against the calculated
pressure drop indicated that the spray header was operating
The scrubber PA section consists of several rows of
multi-pass sheds. The liquid pool at the bottom of the scrubber
vessel is a mixture of cooler pitch feed and condensed heavy
components from the reactor vapors entering the scrubber
through the single cyclone. The pool liquid is pumped and
distributed over the top of the sheds.
As shown in Fig. 4, the vapor from the
cyclone discharges to one side of the vessel, and this
encourages maldistribution into the shed rows. Pressure drop is
generated as the vapor winds its way through the column and
passes through the shed-deck liquid curtains. Low pressure drop
through shed decks exceeds that of packing. Yet, it still
challenges any vapor redistribution.
4. Scrubber snout and shed tray
orientation in plane view.
The directional flow from the cyclone snout, compounded by the
loss of any sheds, could result in vapor-flow channeling. The
performance of the PA return-liquid distributor (H-header in
this case) and uniform cascade of the liquid showers is
consequently critical to maximizing vapor redistribution.
VAPOR DISTRIBUTION FROM SHED PA
Several thermocouples (TIs) are installed in the scrubber
shed section. Readings from the TIs seem to converge at the
time the grid ΔP escalated. Both indications are
located in the Northeastern quadrant. Based on the elevations
under previous normal operating conditions, the 9014 TI reads
vapor temperature and the 9015 TI reads the cooler PA liquid
temperature. After the incident, both TIs converged and yielded
similar readings (Fig. 5). The data indicated
that the TIs are now reading the same phase, either liquid or
vapor, indicating a maldistribution problem.
5. Scrubber shed temperature
profile between TI 9014 and TI 9015.
The TI readings were confirmed using a series of skin
temperatures at four elevations. Cutouts were made in the
insulation so that a thermo-scanning gun could be
used to read skin temperatures at approximately 15°
increments around the vessel circumference.
The skin temperatures had reasonable repeatability, and they
indicated that a hot zone existed at the northwest quadrant.
This position was located at the orientation of the TIs and
confirmed the deviation in readings (Fig.
6. Scrubber shed TI 9014 and TI 9015
from previous turnaround inspection.
The H-header design is a simple, ladder-type slotted-pipe
distributor with laterals positioned over each of the top layer
of sheds. Discussions of the hydrodynamics of liquid flow from
a shed deck are found in literature.1 The up-flowing
vapor can cause atomization of the liquid as it flows through
the curtain, and impingement of the shower on surfaces or other
showers can also cause liquid stream breakup.
Another method was used to evaluate the scrubber shower
liquid throw and was back-checked against internal
geometries to confirm that the cascades from the parallel shed
decks are not impinging on each other.2 The liquid
throw averaged only 15% of the shed spacing (very low weir
loadings). In a perfect design, the H-header liquid would be
distributed proportionally through the H-header according to
the design open area, and each side of the shed should have a
weir loading proportional to the vapor space that the liquid
needed to cover. Shed decks are not ideal mass-transfer
devices, and they are subject to maldistribution (Fig.
7). Accordingly, the actual liquid throw could be
slightly different than the calculated value.
7. Scrubber nozzle and internals
orientations in plane view.
The exit velocity of the H-header end slots is below the target
design guidelines for fouling systems. It is possible that
fouling or coking could have obstructed liquid flow to a
section of the header, resulting in maldistribution of liquid
between the sheds or along the length of an individual shed.
Damage occurring at a lateral or possibly at a flange could
similarly change the header flow hydraulics and alter
distribution. With the low weir loadings, this could result in
one section of the scrubber having very sparse liquid coverage.
Other than the observed temperature deviations, it was not
possible to confirm the H-header performance while the scrubber
was in operation (Fig. 8).
8. Scrubber primary PA distributor
orientation in plane view.
A high C value could contribute to liquid entrainment
from the sheds. The vapor rate to the sheds is approximated by
blending the product stream compositions and estimating the
steam from the reactor side. This does not account for the
heavier components that are condensed within the shed PA
section. However, it should offer a fair approximation of
whether vapor superficial velocity is outside the normal design
Vapor velocity was calculated two ways. The first method
calculates the conventional superficial velocity through the
column cross-sectional area, and the second method uses the
window velocity of the gas.1 The cross-sectional
column area and the window area are very similar in this
column, making the two Cs similar. These Cs
were then compared against the Kister and Olsson Eq. 3 to
determine that this particular shed section operates
significantly away from the flood point.3
Table 2 summarizes the results from these two
The resid entrainment was initially shown to improve with
the use of the startup header, which is an internal open-pipe
tee, positioned above the H-header. When the startup bypass is
in service, the total PA rate and duty increases. Higher duty
results in undercutting GO, thereby improving its quality.
However, the liquid from the startup pipe is poorly distributed
and drops across the central sheds only. If the duty is held
constant, then the side shed TIs in the hot zone are
unresponsive when using the startup header. While using the
startup pipe can increase overall PA flow, it was ineffective
at controlling resid entrainment without major undercutting and
LIQUID/VAPOR DISTRIBUTION INVESTIGATION
After reviewing the equipment design, operating data and
unit impacts to operational adjustments, the factor
contributing to excessive resid entrainment was a malfunction
of the H-header PA return distributor. It reduced liquid
coverage over the northwest quadrant of the column. The liquid
maldistribution promotes vapor maldistribution along with high
localized vapor velocity and Cs by creating a lower
pressure drop pathway, thus leaving the upcoming vapor
ENGINEERING A SECONDARY PA
From using the startup header, it was clear that a
non-targeted application of PA liquid would not resolve the
localized high temperatures. The solution had to direct PA
liquid to the hot zone in a more precise way and be installed
without a unit shutdown.
Several existing nozzles on the scrubber vessel were
identified at several feet above the PA zone. These nozzles
were out of service, and had been part of a decommissioned project to create a spray chamber
that could partially compensate for a fouled grid section
during a previous cycle. This project had used the HKGO wash oil
as a spray media. These existing nozzles were repurposed to
inject PA liquid into the scrubber as a supplement to the
H-header flow. The nozzle orientations around the tower allowed
the liquid flows to be targeted to identified regions that were
observed as hot zones by the temperature surveys.
Engineering proceeded to route PA liquid to the nozzles. The
piping to the new secondary injection location was designed
with the hydraulic capacity to operate as the primary PA,
should H-header liquid distribution problems become more
limiting. A branch was installed on the main PA line, and
routed to a new ring header at the existing nozzle elevation.
The ring header design is fed at two locations and allows for
continuous flow circulation with the flexibility to block and
unblock specific nozzles without creating deadlegs.
Globe valves were added to each arm leading off the ring
header to provide throttling flow to individual nozzles.
Instrumentation was installed to monitor total flow to the
lances, along with pressure and temperature at the H-header and
at each lance location. PA material, by its heavy nature, has
elevated fouling and erosive properties. Accurate, reliable
flowmetering is problematic, and sufficient pressure and
temperature instrumentation was included in the design to
approximate flow as a backup to the flowmeters (Fig.
9. Scrubber primary and secondary
The only means to create back pressure and throttle flow on the
primary line feeding the H-header was an existing block valve
(gate), which has poor control range. A bypass line with a
globe valve was installed with a hot tap around the existing
gate valve to force liquid to the higher elevation where the
ring header was located.
Design of the nozzle inserts
The lances or nozzle inserts were designed at various
lengths, so that each nozzle would be able to target a specific
zone of the vessel by dumping liquid onto a particular shed
deck in the top row. In this manner, the PA liquid could be
applied to the northwest quadrant hot zone to supplement the
H-header poor performance.
An insertable lance would be installed in each of the six
nozzles on the run, avoiding a unit shutdown. The nozzles were
cleared of wall coke using a hot-tap machine, and the lances
inserted through a packing gland. The nozzles were equipped
with a steam purge to be used when oil was not flowing through
the lance to reduce fouling conditions.
The nozzle insert design was intended to accomplish two key
objectives: to allow large droplets of liquid to spray in a
contained zone from an elevation of 9 ft, and to allow
hydrolancing on the run should they become plugged. Several
nozzle designs were bench tested with water to determine the
spray profile, all with high line and exit velocities to
prevent plugging. The best performing design incorporated
attributes for anti-fouling and created a tight cone without
generating mist (Fig. 10).
10. Nozzle design iterations.
Test No. 1: Rectangular slot at end of capped
pipe. The initial flow test showed that the material
retained more forward velocity than expected, despite the end
plate acting to deflect the liquid downward. The pattern was
insufficient to wet sheds directly below the nozzle, making a
hollow arc 4°-ft outward and 8°-ft to each side. This
design would be ineffective in reaching the known hot zone near
the wall when used in the existing nozzles.
Test No. 2: Rectangular slot shifting inward from
end of capped pipe. The exit point was moved further
away from the end cap to create a turbulent chamber above the
nozzle that would direct more flow downward. This test nozzle
showed a similar hollow arc with an even further forward throw
Test No. 3: Internal sleeve. An internal
sleeve was fitted into the exit orifice to force the liquid to
change direction inside the lance, giving the fluid vertical
velocity before it exited the slot. This design showed a more
uniform pattern, centered only one foot outward from plumb.
Test No. 4: Internal weir and sloped end
plate. An internal weir would force the same fluid
directional change inside the lance, and would be easier to
fabricate. The weir covered the full width of the exit slot to
minimize side-throw of the liquid, and the end plate was
modified to slope backward. This design showed a wide oval
cone, but it greatly reduced the forward throw of material.
Test No. 5: Final design. For the final
test nozzle, the slots arc was narrowed to shrink the
wide oval shape of the bulk flow into a tighter full cone. This
design was installed.
Commissioning the secondary PA. The six nozzles and
corresponding lances (A, B, C, D, E and F) were installed, as
shown in Fig. 11. All six lances were inserted
and set on steam purge. Oil was commissioned to the lances that
most closely targeted the hottest areas identified by the skin
temperature scans (nozzles A and F). Once the secondary PA was
commissioned, operating data indicated an almost instant
improvement, and the charge rate was increased without
deterioration in HKGO product quality (Tables
3 and 4).
11. Secondary PA nozzle orientation.
The skin temperature survey was repeated to determine the
impacts on the hot zone from the secondary PA
liquid application. The local temperatures in the northwest
quadrant decreased by 30°F50°F.
Several optimizations of H-header, startup header and lance
nozzles were explored to determine the configuration, resulting
in the most even radial skin temperatures and lowest resid
entrainment into HKGO product. The test with the H-header and
lance nozzles A, F and E in service provided the optimal
results (Tables 5 and 6).
Half of the PA flow was directed to the H-header, while the
other half was directed to the nozzles. Nozzles A and E were
throttled to allow the majority of the flow to be directed to
Although confirmation of the nature of the maldistribution
cannot be made until a future outage, there are several
engineering optimizations that should be considered as lessons
learned in this example.
H-header distributor. The design of the
H-header allows for low velocity at the end of the laterals and
low exit velocity through the slots furthest from centerline.
This is a function of the pipe size and slot area. Redesigning
the H-header with eccentric reducers along the length of the
laterals will allow velocity to be maintained to minimize
fouling. Because this system operates with a low shed weir
loading, orifice sizes must be balanced against velocities
suitable for fouling protection.
De-entrainment grid. There are multiple
options for structured grid/packing styles that may balance
de-entrainment ability against antifouling characteristics
(open area for larger capacity). A limited number of
fluidized-bed cokers are operating that set a reference for
using an alternate grid type.a There may be
opportunity to optimize the bed height and grid type to provide
longer cycle lengths with adequate entrained resid removal.
Fractionator. While the primary issue
caused by resid entrainment is contamination of the HKGO
product, any liquid carried by vapor up to the trays from the
inlet nozzle is subject to coke formation. These trays are
notorious for fouling problems, both with standard coke and
with polymeric coke. Providing adequate means to break the
inlet momentum can prevent liquid entrainment onto the trays,
in addition to optimizing the tray design (or considering
alternatives to trays). HP
This is an upgraded version of a presentation given at the
2013 AIChE Spring National Meeting, San Antonio, Texas, April
a The fluidized coker is the Fluid Coking
processed licensed and owned by ExxonMobil Research and
b FLEXIGRID is a registered trademark owned by Koch
Engineering Co., Inc.
1 Fair, J. R., How to design baffle tray
columns, Hydrocarbon Processing, May 1993,
2 Bolles, W. L., Chapter 14, Design of
Equilibrium Stage Processes, Ed. B. D. Smith. New York,
McGraw-Hill, 1963, p. 498.
3 Kister, H. Z. and M. Olsson, Dont let
baffle tray flood baffle you, the AIChE Spring Meeting,
Distillation Topical Conference,
Chicago, March 2011. Jill Brown Burns is a principal process
engineer at Valero Energy Corp. She is involved in
troubleshooting, operations, and design of the many columns and
crude and vacuum units at the 14 Valero refineries. Her
previous experience included positions at Marathon Petroleum
and Sulzer Chemtech. Ms. Burns has a BS degree in chemical
engineering from the University of Oklahoma.
Daryl Hanson is a technology advisor for
Valero Energy Corp. and is focused on distillation/fractionation/separation
issues. He is responsible for design, troubleshooting,
operation issues at the 14 refineries and Valeros
ethanol plants. His previous
experience included positions at Glitsch, Koch-Glitsch
and Process Consulting Services. Mr. Hanson has a BS in
chemical engineering from Texas A&M University.
Chris Riley is the staff process
engineer at Valeros Benicia refinery. He has supported
Benicia in various technical and operational roles
since hiring on with Valero in 2008. His previous
experience included positions at Chevron and Genentech.
Mr. Riley has a BS degree in chemical engineering from
Brigham Young University.
Cameron Wicklow is the process
engineering manager at Valeros Benicia refinery. He has supported
Benicia in various technical, operational, and planning
roles since hiring on with Exxon in 1999. Mr. Wicklow
earned his BS degree in chemical engineering from the
University of Illinois at Urbana Champaign,