You are on page 1of 6


Revamping FCCs - Process

and Reliability
As FCC revamps push operating severity to higher levels, reliability in the product
recovery section becomes a major issue. In particular, slurry zone coking in the
FCC main fractionators causes unit shutdowns and other problems.
The result: production losses and lower revenue

Scott W Golden
Process Consulting Services Inc

become necessary or very poor prod- Process designs and equipment

he cause of coking is a combi-
nation of temperature and resi- uct quality must be accepted. designs that work with low severity
dence time. In most cases cok- We will present a detailed analysis simply do not work when the unit
ing is not inevitable, it is a conse- of a case where coking was caused by operates with higher temperatures or
quence of design and operating errors. poor process design and operation in lower liquid rates. The use of packing
If the process equipment design is not an FCC main fractionator slurry in refinery main fractionators and low
correct, then many units will coke pumparound section. The problems liquid rate service is common. Using
regardless of how they are operated. If shown in the case history can be packing in FCC main fractionator
the process realities associated with avoided by correct integration of desuperheating section has process,
more severe operation are ignored, the process design, equipment design, efficiency, and mechanical advantages
unit will coke, and a shutdown will installation review, and operations. over shed or disc-and-donut trays.




D/SC &
3.0 M GRID



Figure 1 Process flow scheme Figure 2 Slurry pumparound

Reprinted from PETROLEUM TECHNOLOGYQUARTERLY@ Summer 1996 issue, pgs. 65-93.






Figure 3 Desuperheating section temperatures Figure 4 Column internals

While packing has the advantage of Properly designed FCCU slurry slurry and light cycle oil (LCO) prod-
inherently low hold-up or residence zones can have four-year-plus runs, uct) and the separation efficiency
time, it is much lessforgiving. The engi- without coking. This case study will between thesecuts.
neer must customise the equipment evaluate the theory and operation of Lowering the slurry pumparound
designto the specificsof the system. the desuperheatingsection as it relates heat removal increases the reflux in
An FCCU main fractionator slurry to unquenched liquid temperature the LCO/slurry fractionation section
pumparound section (Figure 1) was leaving the bottom of the grid, grid bed and the temperature of liquid leaving
coking. The column had been efficiency (height of packing), and the grid bed increases.The higher liq-
revamped from disc-and-donut trays requirements of the column internals uid exit temperature from the grid
to grid packing (Figure 2) and the bed to have trouble-free operation. increasesthe quench duty in the bot-
began coking within six months of It will present the specific distributor tom of the column.
start-up. The column had to be shut design error that caused the problem A column operating at a bottom
down to replace the grid bed. and evaluate what caused the coking. pressureof 2.26kg/cmzand high inter-
A seemingly small problem with the The unit has been modified and the nal reflux rates between the LCO/slur-
distributor design caused the bed to most recent unit outage showed that ry requires a high flow rate of quench
coke. The design data supplied to the the grid is coke-free. to maintain bottom temperature. The
equipment manufacturers by a large unquenched liquid leaving the grid
E&C was the typical liquid and vapour Process considerations bed can be 0-39’C higher than the
loadings with no indication of the Figure 3 showsa processsketch of the quenched bottoms temperature.
process flow scheme or composition slurry pumparound section. The col- The high temperature leaving the
differences between the two liquid umn bottom temperature is quenched grid bed was thought to causethe cok-
feeds entering the distributor. The to 360°C by feeding cold circulating ing. In turn, the high temperature was
cause of the problem was a lack of slurry below the bottom liquid level of attributed to having “too much pack-
understandingof how the processand the column. The amount of quench ing”. This is false.
equipment design are linked. Using duty is a function of the column pres- With operating pressure set by
textbook “distillation” designsthat fail sure and slurry composition. The slur- downstreamlimits and composition by
to account for the process specific ry composition is controlled by the product quality specifications,temper-
requirementscausedthe unit failure. column heat balance (reflux between ature is fixed by the process.

Reprinted from PETROLEUM TECHNOLOGY QUARTERLY@ Summer 1996 issue, ~9s. 85-93.

The efficiency of the grid bed affects cold slurry pumparound below the
the slurry bottom product composi- bottom liquid level. The quench
“The operation of the
tion, which changes the unquenched reduces the column bottom tempera-
FCCU desuperheating liquid temperature leaving the grid ture below the slurry product equilibri-
section is sometimes bed. The higher the unquenched liq- um temperature for the column oper-
uid temperature leaving the grid, the ating pressure. The composition of the
/ not well understood in higher the coking potential in this sec- slurry (specifically the quantity of
I terms of its impact on tion of the column. The unquenched 343°C minus material) sets the
unquenched liquid bottoms liquid temperature is 393°C. unquenched liquid temperature leav-
The high vapour temperature, ing the grid for a given reactor effluent
temperature. It is not 527”C, entering the column, and the yield structure. Small changes in the
i a pure heat transfer high liquid temperature, 393”C, leav- quantity of LCO boiling range material
ing the bottom of the slurry in the slurry make large changes in the
i section, where the
pumparound section require a reason- unquenched bottom temperature.
; bottoms unquenched able quantity of liquid in each cross- Figure 3 shows the liquid rates from
/ liquid temperature sectional area of the column to keep the wash section, slurry pumparound,
the bed cool and eliminate coking. If and quench for this example. The
/ leaving the grid goes the liquid distribution to the slurry wash liquid and slurry pumparound
/ up because of better pumparound section is not relatively temperature are 343°C and 238”C,
i heat transfer” uniform across the column cross-sec- respectively.
tional area, or if there is no mixing of The liquid rate leaving the wash
L the lighter wash oil, then sections dry zone is 3720 litres/min and the cold
Determining the cause of the unit fail- out and the grid will coke. slurry circulation is 6620 litres/min.
ure requires a thorough understanding The cold slurry pumparound liquid
of how equipment operates in a partic- Vapour maldistribution represents 65 per cent of the total flow
ular process. The heat and mass trans- Poor distribution of vapour to the to the grid. The slurry pumparound
fer that occurs in the grid bed affects desuperheating section column inter- has much higher heat removal per
the grid bed internal liquid and vapour nals is a reality on all units. The reac- mass flow of liquid than the liquid
temperatures. tor effluent feed to a FCCU main frac- leaving the wash zone because of the
The vapour temperature reduction tionator enters the column on many temperature difference. The slurry is
across the grid comes from heating the units at a velocity in excess of almost 155°C sub-cooled, whereas the
cold circulating slurry to its bubble 45m/sec. The combination of high feed wash liquid temperature is at its bub-
point, and fractionation that occurs velocity and the fact that most units blepoint.
between the liquid leaving the wash have no vapour distribution results in
section and the reactor effluent. It is vapour entering the grid quite maldis- Desuperheating design and opera-
important to understand what hap- tributed. The liquid distributor design tion
pens to the process stream tempera- controls the quality of liquid distribu- The operation of the FCCU desuper-
tures entering and leaving the grid bed tion. With the inherently poor vapour heating section is sometimes not well
(Figure 3) to evaluate the potential distribution and large quantity of understood in terms of its impact on
causes of coking. The vapour tempera- desuperheating that takes place, it is unquenched liquid temperature. It is
ture decrease across the grid bed is possible that dry sections of the pack- not a pure heat transfer section, where
approximately 178°C. ing can result without adequate liquid the bottoms unquenched liquid tem-
The liquid temperatures entering the distribution. Slurry grid bed coking is perature leaving the grid goes up
grid for the cold circulating slurry and largely a function of proper distribu- because of better heat transfer.
the liquid leaving the wash section are tion of both the liquid leaving the The efficiency of this section of the
approximately 238°C and 343”C, wash section and the cold circulating column affects the unquenched liquid
respectively. The flow rates of the indi- slurry. temperature by changing the amount of
vidual liquid streams entering the grid light ends (relatively speaking) in the
bed are a function of the column heat Bottoms quench slurry product. When increasing the
balance (for a given reactor effluent). Bottoms quench is employed to avoid efficiency of this section of the column,
The combined liquid temperature coking problems in the bottom of the for given heat balance, the unquenched
entering the grid bed is 271°C for this fractionator vessel. The main fraction- liquid temperature will go up.
case. Hence, a large quantity of the wash ator bottom temperature is controlled If a desuperheating section grid bed
liquid is revapourised in the grid bed. in the range of 360-370-C by feeding has a poor quality liquid distributor

Reprinted from PETROLEUM TECHNOLOGY QUARTERLY@ Summer 1996 issue, pgs. 65-93

and it is replaced with a better quality could not meet the duty requirement.
distributor, the unquenched liquid If the heat transfer requirement is
“We’ve seen some
temperature goes up because the slur- not met in the desuperheating section,
ry product lightends content goes the upper section of the column will models using three to
down. This has been proven on a unit have to desuperheat the vapour and five stages for this
where the slurry product flashpoint the column will probably flood. In
increased from 57°C to 9 1 “C by chang- addition, the grid will probably coke.
section of the column.
ing only the liquid distributor. There Never size the packed bed volume This is wrong. The use
was no heat balance change on the in the slurry zone of an FCCU with a of two theoretical
unit, and the required quench flow coefficient higher than 8,00Ow/mJ-“C,
rate increased. or problems can result. stages in the model
If the height of packing in the desu- The equation used to calculate the will always yield
perheating section or the efficiency of heat transfer coefficient for a grid bed
the packing is increased, the is based on the equations shown
higher vapour/liquid
unquenched liquid temperature will go below: loading than reality.
up because the lightends content goes U = 1019 Fl.6 ( &) 0.96 The result of these
down. The equilibrium bottoms tem-
perature is set by slurry composition theoretical issues is
and column pressure. As in any equi- F =X--L that the vapour/liquid
A, PI-Pv
librium situation, the composition,
loadings used to size
pressure, and temperature are related. where
You cannot change one variable with- U = heat transfer coefficient in
the beds are usually
out a corresponding shift in the others. w/m3-“C higher than reality.”
Normally, these grid beds are sized L = liquid rate in litres/min -
by calculating the volume of packing A, = column cross-sectional area, m2
based on a heat transfer equation. The V = vapour rate, mJ/sec cal stages in the model will always
equation is: yield higher vapour/liquid loading
Q= U V LMTD The liquid and vapour rates used in than reality.
where the equation are based on the highest The result of these theoretical issues
Q= heat transferred loadings in the bed. In a desuperheat- is that the vapour/liquid loadings used
Ll = heat transfer coefficient based ing section, the highest vapour/liquid to size the beds are usually higher
on packing volume loads are in the middle of the packed than reality. In addition, the grid
V = volume of packing bed. The so-called bubble effect, equations were derived in the late
LMTD = log mean temperature dif- which occurs in the middle of the 1960’s and it is questionable whether
ference. bed, is calculated from a computer the data used were sufficiently accu-
The packed bed volume is selected model. rate.
based on the quantity of heat that The model must use at least two In this case, where the bed coked,
must be transferred. theoretical stages to generate this bub- the calculated heat transfer coefficient
Unlike most refinery tower heavy oil ble (with one stage the models will not was stated to be 17,750w/m3-“C, based
pumparound packed sections that pro- generate the bubble). Keep in mind on applying the equation. The bed vol-
duced heat transfer coefficients in the that a theoretical stage of efficiency ume installed in the column was more
range of 8,000-16,000w/m3-“C, the zone is approximately 2.4m. A 3m bed than twice that required, assuming the
actual heat transfer coefficients mea- does not have two theoretical stages. calculated heat transfer coefficient was
sured for the FCC slurry pumparound The computer models generate a high- correct. When the bed coked, the
(as best one can in the field) are more er bubble (higher loadings) as the problem was attributed to too much
like 5000w/mJ-“C. In other words, if number of stages used to simulate the heat transfer, causing high
setting the bed volume based on a desuperheating section is increased. unquenched liquid temperature leav-
higher heat transfer coefficient than As the number of stages used to simu- ing the bed.
can be achieved in actual operation, late the slurry pumparound section One analysis of the desuperheating
there is a high potential that the bed increases, a dramatic rise in the calcu- section based on pure heat transfer
will not meet the heat transfer require- lated loadings will be realised. We’ve predicted the bottoms temperature
ment. seen some models using three to five would be 427°C. To reach 427°C slurry
In at least one case, the grid was stages for this section of the column. product equilibrium temperature, the
removed because the bed volume This is wrong. The use of two theoreti- LCO product would have to be yielded

Reprinted from PETROLEUM TECHNOLOGY QUARTERLYa Summer 1996 issue, pgs. 65-93.

to a 482°C endpoint and the initial However, accurate field data from a Figure 3 shows the liquid rates
boiling point of the slurry product similar unit showed the actual heat entering the parting box based on the
would be close to 343°C. The transfer coefficient was approximately column heat and material balance. If
unquenched bottoms temperature is a 7,OOOw/m3-“C. By using the lower heat one makes an assumption that half of
function of the bed height, but it is the transfer coefficient, the minimum grid the distributed wash liquid to the grid
fractionation efficiency of the bed that bed depth was calculated to be bed has no cold slurry pumparound in
controls the product quality. The light- approximately 3.0m. In comparison, it, and only wash liquid contacts the
ends content sets the un-quenched the vendor supplied the refiner with a reactor effluent, then what would be
bottoms temperature. calculated heat transfer coefficient the results? Based on a simulation of
based on an equation that gives a very maldistribution for this unit, almost all
Revamp optimistic heat transfer rate. the wash liquid is revapourised. What
Figure 4 is a schematic showing the The actual heat transfer coefficient happens in the column is much more
revamped column internals. The liq- from the bed is difficult to calculate on complex. Nevertheless, it is apparent
uid is distributed to the packed bed any unit. An accurate set of tempera- that the grid bed will have some dry
with a gravity flow distributor consist- ture data around the bed was not avail- sections or at the very least sections
ing of two parting boxes and several able. No additional field checks, like without much liquid.
troughs. The parting boxes take liquid radial temperature profiles above and The conclusion was that the column
feed from the bottom wash zone tray below the bed, were used to get better was coking because the liquid leaving
seal pan and feed piping that distrib- temperature data. The assessment of the wash section, and the cold slurry
utes the cold slurry pumparound. the problem was based solely on a pumparound, were not being mixed.
The wash liquid was fed to one end design heat transfer coefficient of The lack of mixing resulted in “light”
of the parting boxes through two weirs 17,75Ow/m3-“C provided by the vendor. hot liquid, leaving the wash section,
cut in the seal pan. The cold slurry liq- The liquid distribution system was being preferentially distributed to a sec-
uid was fed to the remaining section of not considered integral to the success tion of the grid bed. Most of the liquid
the parting box. The slurry of the revamp. As with many revamps, in this zone of the bed revapourised,
pumparound piping fed liquid to only the distribution systems are consid- causing the grid to coke. The liquid
about two-thirds of the parting boxes ered secondary to such issues as distributor caused the coking and it
length. packing HETPs that are usually deter- needed to be modified. (Figure 5)
The liquid distributor was preferen- mined in small diameter test columns.
tially feeding the liquid from the wash Column internal modifications
section to one end of the paring boxes Cause of coking - no liquid mixing The column internals were modified to
and cold slurry to the other. The liquid Figure 4 shows the internal column properly mix the liquid leaving the wash
leaving the wash section and the cold arrangements for feeding liquid to the section and the cold circulating slurry.
slurry were not mixed. The grid bed distributor. Analysing the problem, (Figure 6) The bed depth was reduced
pressure drop began to increase about assuming there is no mixing of some from 3.0m to 2.4m to allow space for
six months after the initial startup. portion of the wash liquid and the removal of the grid bed without having
The pressure drop continued to slurry pumparound, highlights the to remove the liquid distributor.
increase until the unit had to be shut cause of the problem. The nearest column man-way was
down after two years to have the grid There is some mixing of these above the wash section. The space
replaced. streams where the liquid leaving the between the distributor and the pack-
wash section and the cold slurry ing allowed the packing to be removed
Problem analysis pumparound liquid meet in the part- with the distributor in place. The liq-
The column operating data and the ing boxes. However, some portion of uid distributor was modified to proper-
column internal design was reviewed. this liquid leaving the desuperheating ly mix the liquid in the parting box
The refiner’s central engineering section distributor troughs is only liq- prior to distribution into the troughs.
department evaluated the column’s uid from the wash section. The grid bed now received a liquid of
packed bed design and determined This should be expected. Liquid res- uniform composition and temperature
that the problem was too much grid. idence times in distribution parting across the entire column cross-sec-
The calculated grid bed heat transfer boxes operating at moderate liquid tional area.
from the sizing equation used by the rates (typical for pumparound heat Inspection of the coked packing
vendor showed the minimum grid bed transfer zones) are of the order of one before modifying the liquid distributor
depth was I.2m, using the calculated to five seconds. Obviously, the prob- system showed the column was coked
heat transfer coefficient of lem is aggravated as tower diameters in the area under the distributor where
17,5OOw/m3-“C. are increased. there was only wash liquid being fed.

Reprinted from PETROLEUM TECHNO!-OGY QUARTERLY@ Summer 1996 issue, pgs. 85-93.

The section of the column receiving the Conclusion technology that work at lower severity
cold circulating slurry was coke-free. Refinery main fractionator operating cause coking in FCC columns operat-
The cause of coking in this column was severities are increasing due to the ing in a more severe mode.
not too high a grid bed depth. The cause economics associated with these Unit reliability can be improved
of the problem was poor liquid mixing. process operating changes. The inci- while operating with higher unit sever-
The issue of liquid mixing must be dence of unscheduled unit shutdowns ities if proper allowance is made for
considered in refinery packed column to replace coked column internals is integrating process design, equipment
revamps or problems can result. increasing. Coking results from design, installation, and operations.
Liquid mixing prior to distribution to a process and equipment design errors.
packed bed is very important. In this Correcting process design errors and
case, composition gradients due to no understanding the operating variables Scott W Golden is a chemicalengineer
mixing resulted in coking. The bed that must be monitored and controlled with Process Consulting Services, Inc.
pressure drop has not increased since will prevent coking. Ignoring the reali- Houston, Texas. The company provides
start-up. During a unit outage after the ties will not change the inherent revamp and field trouble-shooting for
distributor modifications were made, nature of these refinery processes. the refinery industry worldwide.
the bed was inspected and it was Field operating performance has
found to be free of coke. shown that design procedures and






_._. __ _-^--_. - __.-..----- __-- -..___.--_-___

Figure 5 Localized Coke Area Figure 6 Parting Box Modifications


Suite 130
@ [l]-(713)-665-7046
El [l]-(713)-665-7246

Reprinted from PETROLEUM TECHNOLOGY QUARTERLY@ Summer 1996 issue, pgs. 65-93