You are on page 1of 8

D E B OT T L E N E C K I N G

Uncover Low-Cost
Debottlenecking
Opportunities
he recent strong domestic econ- nuggets often lie. Furthermore, we will

Don’t immediately
concentrate on a
T omy has presented many sec-
tors of the chemical process in-
dustries with a problem every
plant loves to have — how to make more
product. Too often, though, companies
discuss ways that allow the processing
of larger feed streams without necessari-
ly having to increase the internal
throughput in equipment. When this ap-
proach is successful, it can produce
specific piece of do not thoroughly understand an existing more pounds of product at the end of the
plant’s capacities and, so, are not pre- year with minimum capital investment.
equipment. Instead, pared to challenge the existing facility to
produce more. These companies take the Getting started
look at the overall “quick and easy” way to produce more Understanding the existing plant ca-
product by spending capital to duplicate pacity is a critical first step; representing
process and the equipment or even entire processes. Or, it mathematically (at this point, a simple
worse yet, they do nothing and com- material balance) may be the easiest way
interactions among pletely miss a business opportunity. to get there. Ideally, knowing the real ca-
Sometimes, spending major capital is pacity (not just what the design book
units. This often the only way to achieve a business goal said 20 years ago when the plant was
but, many times, the existing plant has built) before an expansion opportunity
can identify the hidden capacity that can be tapped more presents itself will save a lot of time.
most-cost-effective quickly and at less cost. This may be the difference between a
The literature is rife with articles that successful or a failed plant response to a
solution. discuss debottlenecking specific equip-
ment. All these references tout the tradi-
potentially fleeting business opportunity.
Keeping this understanding up-to-date
tional approach of “how can one get should be an ongoing activity of the
more hydraulic capacity through an ex- plant’s technical and operating support
isting piece of equipment.” To this end, staff.
Dave B. Litzen,
the classical approach for distillation, as Often, the modeling step is the most
Virtual Ideality
an example, is to revamp columns with intimidating. Pure-component physical
José L. Bravo,
internals that allow more traffic without properties, binary interactions, azeotropes
Equilon Enterprises LLC
a loss in efficiency rather than to im- — scary, isn’t it? The initial representa-
prove the workings of the installed tion, however, does not necessarily have
equipment; some recent articles do look to be a full-blown, rigorous simulation
at squeezing more out of existing hard- model. For example: An existing oxy-
ware in a column, as well (1, 2). genated-hydrocarbon plant (producing
This article instead concentrates on four finished products in shared equip-
possible synergy among existing equip- ment) was to be debottlenecked several
ment in the entire plant. We have found years ago. Assembling a spreadsheet to
that it is within the interaction of all the back-calculate flow rates to individual
process steps and equipment where the equipment or sections (mainly distillation

CHEMICAL ENGINEERING PROGRESS • MARCH 1999 ©Copyright 1998 American Institute of Chemical Engineers. All rights reserved. Copying and downloading permitted with restrictions.
D E B OT T L E N E C K I N G

columns and reactors) based on the Armed with this predictive tool, model representing this process, and
product flow rates was enough defini- you now should document each that we’ve determined all the major
tion to generate several debottleneck- major piece of equipment’s maxi- unit operations’ maximum feed rates
ing plans over the years. Due to the mum-achievable feed rate and make (marked by an “X” on the streams in
discontinuous nature of this plant, the sure it is consistent with the one in Figure 1). The debottlenecking tar-
spreadsheet was a very handy tool to the material balance. (The term get for this example is 255 million
predict impacts on shared equipment. “major” has no strict interpretation. lb/y of finished product ABC.
The evolution of process simula- Although our example evaluation Once the maximum equipment
tion software has made producing a leaves out pumps, control valves, fil- feed rates have been determined,
simple material balance relatively ters, and so on, these types of equip- combining these data with the pre-
easy. At this point, it is not necessary ment can be included, though at the dictive tool completes the “ever-
to rigorously develop special physi- risk of increasing the complexity of green” process of maintaining plant
cal properties. A total plant flow- the study.) Design data are always capacity knowledge. We’ve observed
sheet simply representing all the useful here but not necessary. The that using the sensitivity analysis
major unit operations with flow real limiting flow rates often can be feature found in some commercial
splitters, mixers, and component determined by examining the plant’s simulators works very well for this
separators (no flashes yet) is the best historical process data. If these data activity. By generating plots for each
way to get started with the material are available electronically, finding individual unit operation (unit-oper-
balance. Using simulation software the maxima statistically is very use- ation feed rates vs. plant finished-
to complete the material balance will ful. This does not necessarily deter- product rate shown in Figure 2), you
allow you to more accurately trans- mine the true equipment capacity, can transpose the equipment feed-
late finished-product flow rates into just the maximum flow that the unit rate limits on each plot to determine
equipment feed rates, which are im- has ever successfully handled. the overall plant product rate at
portant for the steps described In addition, interview the plant which that equipment will become a
below. Also, if you are interested in operators for their thoughts on maxi- bottleneck. In this example, the max-
eventually producing a rigorous mum flows. The operators may well imum feed capacity for the stripper
flowsheet model, working up a sim- provide some insight on maximum before flooding occurred (or will
ple material balance in a simulator is feed rates to each piece of equip- occur, if this is an estimated limit vs.
a very good way to get the rigorous ment, along with some ideas of what a limit based on plant data) is 117
modeling activity started. is leading to a feed limitation. For gal/min, and for the extractor is 5.8
A reasonable plant material-bal- example, you may find that a distil- gal/min. These feed rates, provided
ance flowsheet usually can be assem- lation column is designed to handle all other bottlenecks were removed,
bled in a matter of days. This exer- 120-gal/min feed, but the operators would translate to an annual ABC
cise will force you to answer certain will not run the column at more than production rate of 210 million and
essential questions like what are ac- 100 gal/min, often for very good rea- 218 million lb/y, respectively. Obvi-
tual stream flows and compositions sons. The feed limit to this distilla- ously, this plant will have to imple-
— details necessary for a fundamen- tion column may not be a mechani- ment a capital improvement or, bet-
tal material balance and good for fu- cal limit, but it is a real limit and ter yet, a low-cost process change to
ture rigorous model development. needs to be understood. If no reliable remove these bottlenecks to achieve
One caveat in using a simplified data are available for a unit’s maxi- the target rate of 255 million lb/y.
model to represent the plant is that the mum operating capacity, you will
model will not cover impacts related have to calculate the capacity. Esti- The stair-step chart
to vapor/liquid equilibria (VLE), such mating the capacity of a distillation Once you have translated all the
as compositional changes due to vary- column, for instance, is done fairly major-equipment-limiting feed rates
ing conditions over a range of produc- easily with a commercial simulator to plant product rates, summarizing
tion rates. These effects often are (which usually gives a conservative the bottlenecks on a “stair-step”
minor, but you must apply some engi- value) or a standalone tray-capacity chart by arranging them from lowest
neering judgment to determine the program. to highest appears to be the most ef-
level of model complexity required to To illustrate the process of assem- fective way to communicate the data
best represent your process. Adding bling a debottlenecking plan, let’s (see Figure 3). Technical and non-
controls (fixing ratios and stream use an imaginary process that pro- technical people alike seem to quick-
composition targets, for example) can duces a product called ABC. Figure ly grasp a presentation of data in this
compensate for some of the thermo- 1 depicts the simplified process single-page form.
dynamic shortcomings of the simple schematic. At this point, we will as- To enhance the usefulness of the
model. sume that we have a simple process stair-step chart, add cost estimates

CHEMICAL ENGINEERING PROGRESS • MARCH 1999


■ Figure 1.
Simple schematic
of the ABC To Downstream Unit
process.
Reactor Flasher
Reactant A Stripper
Reactant B X
Reactant C Preheater Steam
X X
X

To Downstream Unit

Solvent ABC Process


Flowsheet
Extractor
X
Finished ABC

6 6
5 5
4 Crude 4 Finishing
X Column X Column
3 3
2 2
1 1

Byproduct Purge

for debottlenecking each piece of the stripper capacity is expanded, the play for this evaluation (unless its
equipment; this allows easy visual- new stripper-capacity bar can be new capacity can be utilized some-
ization of the cost/benefit of each moved toward the right on the chart how to remove other bottlenecks
“step” as you progress toward your and reranked against all the other with simple low-cost process
debottlenecking goal. For this cost unit operations. As long as the ca- changes). Now, the extractor be-
estimate, assuming that the equip- pacity is above the target line, the comes the plant bottleneck. In
ment bottleneck cannot be removed stripper no longer will come into essence, the $2-million expenditure
by a process change will result in the
most conservative estimate. This
chart also will identify the equipment
that has more-than-adequate capacity 150 10.0
to achieve your goal. Having this in- 145 Stripper Feed 9.5
formation gives you the opportunity 140 Extractor Feed 9.0
to creatively remove an equipment
Extractor Feed, gal/min
Stripper Feed, gal/min

135 8.5
constraint below the target produc-
tion rate by taking advantage of extra 130 8.0
equipment capacity elsewhere — that 125 7.5
is, equipment synergy. 120 7.0
The chart in Figure 3 indicates
115 6.5
that the stripper, as mentioned previ-
ously, is the primary plant bottle- 110 6.0
neck. The example presumes that for 105 5.5
$2 million the plant can raise the ca- 100 5.0
pacity of the stripper to support plant 200 210 218 225 250 275 300
rates of 255 million lb/y or more. Annual Production, million lb/y
Figure 2 can be used to determine
the minimum feed rate to the strip-
per at the debottlenecking target (see ■ Figure 2. Plotting individual-equipment feed rates vs. annual production can
the dotted extrapolation line). After pinpoint the amount of additional feed required to meet the output target.

CHEMICAL ENGINEERING PROGRESS • MARCH 1999


D E B OT T L E N E C K I N G

350

300

Debottlenecking Target
250

200

150
Cost to
Debottleneck
Above Target
100
Finishing
Stripper Extractor Column Reactor Crude Column Flasher Preheater

Capital Cost, $ million 2 0.5 1 0.5 0 N/A N/A


ABC Production, million lb/y 210 218 235 251 274 300 310

■ Figure 3. A single stair-step diagram clearly shows which major equipment need to be debottlenecked, and which have excess
capacity that may provide debottlenecking opportunities.

for the stripper increased overall debottlenecking target is no more capacity enough to reach the re-
plant capacity by 8 million lb/y, than 20–30% above the nameplate ca- moval target. Or, taking advantage of
which as such may be a poor invest- pacity, you should have a pretty good the extra capacity in the crude col-
ment if that were the only step taken chance of utilizing the existing plant umn by raising reflux to improve
toward the 255-million-lb/y capacity (due to safety factors and conserva- separation might allow decreasing
goal. tive assumptions used in its design) to the solvent rate to the extractor, re-
Generally, all the unit operations achieve your debottlenecking goal. ducing reflux on the finishing col-
that fall below the target line need to So, how can an evaluation like umn and total flow to the reactor.
be expanded to achieve the goal. In this save money? By pinpointing This should result in increased ca-
some situations, however, the step- which equipment require debottle- pacity for all three unit operations
wise removal of all bottlenecks be- necking and which do not, it allows with one small process change in the
neath the target line might result in a you to focus on possible synergism. crude column.
case of diminishing returns. The data For instance, Figure 3 shows that the This systematized approach to
on the stair-step chart provide all the crude column, the flasher, and the understanding plant capacity is most
basics for a cost-benefit analysis. For preheater have capacities greater useful if done before the need to de-
example, if the cost to raise the reac- than the debottlenecking target, bottleneck presents itself. Ideally,
tor capacity were $5 million instead while the stripper is the primary bot- once the stair-step chart has been as-
of $0.5 million, it might make sense tleneck fixing the present plant ca- sembled, you should keep it ever-
to lower the target to the current re- pacity. These observations should green by updating it once per year or
actor capacity. If the target is far stimulate some potentially-low-cost whenever a major piece of equip-
above the capacity of most unit oper- debottlenecking solutions. One idea: ment is replaced or added.
ations in the plant, the cumulative increasing the duty of the preheater
debottlenecking cost from the stair- and removing more volatile com- Generating, evaluating,
step chart readily might make appar- pounds in the flasher (as both have and ranking ideas
ent that building a new plant is the extra capacity) would reduce the re- Once the overall plant capacity
only cost-effective solution. quired steam-stripping load in the has been assessed, you can turn to
Our experience suggests that if the stripper, which might boost stripper the fun and exciting part of being

CHEMICAL ENGINEERING PROGRESS • MARCH 1999


creative to remove bottlenecks. The fixing model input and comparing enue), earning power (maximum
game (okay, objective) is to come up calculated results. Validating the cost of capital at breakeven point),
with ideas that are based on process model at different rates and condi- or capital efficiency (5-y net present
changes that require no capital. Use tions would result in a more robust value divided by capital cost). Pro-
the stair-step chart to pinpoint where model, but is not really necessary for cess risk and staff time consumption
the priorities are and where the syn- evaluating debottlenecking options. also may require consideration in the
ergism might lie. It’s almost a neces- Using any simulator, models can ranking process.
sity at this point to develop for the be run manually to adjust process
major process steps rigorous simula- parameters (column tray efficiencies Some real-life examples
tion models (ones that use funda- and heat-transfer coefficients, to We already have used this strate-
mental physical properties and pos- name two) to match plant data. To gy successfully on numerous pro-
sibly additional interaction data to aid the validation process, some sim- jects, including:
determine the VLE and VLLE) to ulators have tools to regress process Reactor/column combination. In
help evaluate your creativity. Having parameters using techniques similar the oxygenated hydrocarbon process
validated, rigorous models not only mentioned previously, the plant was
will improve the accuracy of the re- asked to exceed its understood ca-
sults but also the speed at which they pacity for product. The stair-step
are generated. If process models do Use the stair-step chart indicated that the reactor was
not already exist, generating new limiting the manufacturing process.
ones may be the most-time-consum- chart to pinpoint The initial impulse was to install an
ing step in the entire evaluation pro- additional reactor. But, with the
cess — particularly if component where the priorities knowledge we had of the equipment
physical properties are not immedi- constraints, we knew that the purifi-
ately available for your process. are and where the cation column downstream of the re-
Having a proven set of physical actor had plenty of capacity. We also
properties for process modeling is synergism might lie. learned from operator interviews that
important for all processes, but ex- the reactant concentration target in
tremely important for those that may the reactor product, set at 0.1 wt. %
contain many nonideal interactions. to meet finished product specifica-
Indeed, your success may well de- to those for physical-property regres- tions, was what established the max-
pend on an accurate representation sion. Additionally, these tools usual- imum feed rate to the reactor. As an
of these physical properties. ly can help to reconcile plant flow- alternative to installing a new reac-
Let’s assume the component phys- rate data (as raw plant flow-rate data tor, the plant instead relaxed the self-
ical properties are available for your rarely do material balance) to im- imposed reactant concentration in
process. Of course, having process prove validation. the reactor exit to 1 wt. %, allowing
models that mimic the plant is the de- The validated models now can be an increase in feed to the reactor.
sired objective. Model convergence used to evaluate debottlenecking op- Using the excess capacity of the pu-
does not guarantee this (3). To prop- tions case by case. The simulator’s rification column, the plant boosted
erly validate a process model, you sensitivity analysis again can be a reflux, top product recycle, and con-
must compare the model input and re- time-saving feature for the case trol tray temperature (consistent with
sults against stable plant data. Analyt- evaluations. model predictions) to ensure finished
ical and process data that are normal- The evaluations hopefully should product specifications were met.
ly recorded to operate the plant often result in many projects, both capital Sounds simple — but the plant was
are adequate to use for model valida- and noncapital. These should be prepared to spend $1–2 million to
tion. Many modern plants have these ranked, with the noncapital projects achieve the same results. The alter-
data electronically archived for easy given immediate consideration for native was free, except for the minor
access and evaluation. implementation. In addition, estimat- variable cost due to higher recycle
To generate a basis for the valida- ing the installed capital cost for the and reflux. More importantly, the re-
tion, rely on data taken during stable remaining projects is necessary. Fol- sults were immediate.
plant operation — ideally 30–50 lowing that, you must choose an Column pressure leveraging. In
consecutive days with no upsets or economic index and index target another oxygenated-hydrocarbon
interruptions — to determine pro- consistent with business manage- manufacturing plant, the finishing
cess means (or medians, if data are ment objectives. Examples of an column was designed to remove and
not normally distributed). Then, use economic index are simple payout recycle reactant (a higher boiling
the resulting means (or medians) for (capital cost divided by annual rev- component than the finished product).

v
D E B OT T L E N E C K I N G

The 100-tray column required a rela-


tively high reflux to successfully puri-
fy the finished product to its required 1.10
specification. This, in turn, consumed 1.05
much of the column capacity. The

Flood Factor
column, for convenience, was operat- 1.00 Flood Factor
ed at atmospheric pressure. 0.95
Very often, changing the column
0.90
pressure can squeeze a few percent
more capacity from distillation 0.85
columns. Decreasing the system
0.80
pressure usually results in increased 4 6 8 10 12 14 16 16.7 18 20
relative volatility of the key separa- Pressure, psia
tion components, reducing the re-
quired reflux (or number of stages)
to meet a separation target. On the ■ Figure 4. With fixed top and bottom compositions, a distillation column model can
other hand, reducing the pressure de- exhibit a minima when plotting flooding factor vs. operating pressure.
creases the vapor density, increasing
vapor velocity and cutting overall
column capacity. Because of these is by reducing pressure and lowering ty. In the case of the DIB, however,
two counteractive effects, a natural reflux. In the retrofit of a debutaniz- the gains from improved separation
pressure optimum will occur (1). er and deisobutanizer (DIB) combi- performance from more trays at
This effect is easily observed by as- nation in a gas plant, we went even a lower pressure overcome the loss of
sembling a sensitivity analysis in a step further by decreasing tray spac- hydraulic capacity by requiring
column simulation. ing (a step traditionally associated lower reflux. When this is combined
For the oxygenated-hydrocarbon with lower capacity) to obtain more with the superior flooding capacity
finishing column, we can fix the stages and lower reflux. of the improved sieve trays, the DIB
composition targets of the key com- Three issues turned out to be can be debottlenecked by moving in
ponents (the product and the reactant most important for the integration the seemingly counterintuitive direc-
in this case) in the top and the bot- and debottlenecking of these two tion of lower tray spacing (more
tom of the column, vary the column towers: trays) and lower pressure (better
operating pressure independently, 1. Increasing debutanizer pressure volatility).
and change the reflux and purge rate — to provide hotter overhead that Feed-port/sidedraw optimization.
to meet the target compositions. can be used to reboil DIB. Many distillation specifications tar-
Have the simulator estimate the 2. Decreasing DIB pressure — to get very low concentrations of sin-
maximum flood factor for each pres- give cooler bottoms that can be re- gle, identifiable compounds. These
sure case. Plotting flood factor vs. boiled by the debutanizer overhead, stringent targets often can lead to
pressure across the range of evaluat- raise relative volatility of isobutane opportunities to optimize feed-port
ed pressures allows you to observe (thus enabling lower reflux), and cut and, if applicable, sidedraw loca-
the optimum (see Figure 4). hydraulic capacity available in tower tions on purification columns. For
For the oxygenated-hydrocarbon shell. example, a chlorinated-hydrocarbon
finishing column, a 1–2% capacity 3. Adding to DIB tray count — to finishing column (with a sidedraw
increase should be achieved by mov- boost theoretical-stage count and re- product) was designed to remove
ing the operating pressure up slightly duce reflux required, at the expense of both heavy and light chlorinated im-
from atmospheric to 16 psia, the op- tray spacing and hydraulic capacity. purities from the finished product.
timum. In the actual case, more im- The combination of items 2 and 3 Reflux ratio already was maximized
provement was required; so, the col- results in large energy savings and to meet purity specifications and re-
umn was refitted with higher-capaci- increased capacity due to lower in- duce yield loss. The plant was unable
ty, more-efficient trays to obtain the ternal column traffic. Additional hy- to incrementally increase feed rate
target capacity. draulic capacity in the DIB at re- without sacrificing either purity or
Column pressure leveraging — duced pressure and tray spacing only yield loss. Using a sensitivity analy-
the other way around. Sometimes, can be achieved via high-capacity sis similar to that described above,
decreasing the pressure of a column sieve trays. Interestingly, pressure optimum feed-port and sidedraw lo-
can remove a bottleneck. Indeed, a and tray-spacing reduction normally cations can be determined simultane-
classic way to increase feed capacity are considered to cut column capaci- ously by maximizing vapor traffic

CHEMICAL ENGINEERING PROGRESS • MARCH 1999


replacing the furnace burner tip with
0.99970 a new design (about $1,000 in-
stalled) that enabled the preheater to
0.99965
operate with 30% more heating duty.
0.99960 During the occasional low-product-
Product Purity

0.99955 S/D Tr2 S/D Tr6 rate periods where the fired pre-
0.99950 S/D Tr3 S/D Tr7 heater becomes the bottleneck, yield
S/D Tr4 S/D Tr8 loss has been reduced, saving the
0.99945 S/D Tr5 X Base Case-S/D Tr13
plant up to $1 million/y.
0.99940 X Use of flooding detection mecha-
0.99935 nisms. One way to maximize capaci-
25 26 27 28 29 30 31 32 33 34 35 36 37 38 ty in a process is to maximize the ef-
Feed Tray Location fective onstream factor. Too often,
though, plants push their equipment
without considering the impact on
■ Figure 5. A sensitivity analysis can be used to determine optimum feed-port and onstream factor. Yet, when processes
sidedraw locations. are operated very close to their lim-
its, excedences can mean lengthy re-
covery periods that end up reducing
(by maximizing reflux at a given feed light ends to a very low concentra- the real onstream factor and output.
rate), fixing the product loss in the tion to meet the finished product This is particularly true in distilla-
purge stream, and varying feed-port specification. The second light-ends tion steps — a column that floods
and sidedraw locations to locate the column was the smaller of the two, may not produce valuable product
point of highest purity (see Figure 5). and limited the capacity of the over- until it completely recovers from the
You can use other combinations of all finishing process. Reevaluating flooding episode.
variables (fixing purity and yield loss the column capacities individually If pushing a column to 98% of its
to determine minimum reflux, for in- indicated that each column on its capacity from 90% lowers reliability
stance) to gain the same results. own could meet the light-ends speci- to, say, 90% from 100%, you actually
In both this and the previous pres- fication. So, the plant has been mod- wind up worse off, because output will
sure-optimum case, a flowsheet opti- ified (via piping changes) to operate have dropped to 88% of nameplate
mizer (found in many flowsheet sim- the light-ends columns in parallel in- from 90%. The trick is to operate at,
ulators) also could be used. This op- stead of in series, providing a quick, say, 98% of capacity 100% of the
timizer should come up with the 40% increase in light-ends removal time. That would represent removal of
same answer as the sensitivity analy- capacity for minimal cost. an 8% bottleneck in the operation.
sis, but only will give the optimum Compressor/preheater balancing. For columns, the use of reliable
point (if one exists). We prefer the In a plant with a light olefin feed- indicators to warn the operator that a
sensitivity analysis, because it helps stock, the process requires vapor re- flood episode is approaching may
visualize the optimum. This can be compression and preheating of the prove very valuable. If the column is
very valuable in determining whether olefin before it is reintroduced to the at 98% of maximum capacity and an
the independent variable optimum reactor system. The stair-step dia- incipient-flooding indicator turns on,
has a flat or sharp response and, thus, gram indicated that the compressor the operator may choose to back off
how sophisticated solutions need to was the bottleneck to increasing pro- the feed to avoid the costly flood
be to maintain the optimum. In this duction capacity at the present episode. Then, once the danger is
case, the true optimum is in a posi- byproduct yield, but the fired pre- averted, the capacity of the column
tion (feed port on Tray 27 and side- heater was the bottleneck if the plant can be challenged again, using the
draw on Tray 2) that does not exist. was targeting reduction of yield loss flooding indicator as a warning.
The shape of the curves, though, led at a fixed capacity. Simply under- By using such indicators, we suc-
to the best choice of existing feed- standing the limitations, tradeoffs, ceeded in debottlenecking a large
port and sidedraw location, a change and synergies between these two distillation column that had some re-
that was made for no capital cost and pieces of equipment allowed the liability issues at high rates. The ef-
minimal loss of production. plant to increase product rates by fective capacity of the column was
Column realignment. In the chlo- 15% by operating closer to the com- increased by more than 5% by oper-
rinated hydrocarbon system men- pressor capacity threshold. The ating the column at a rate equivalent
tioned above, the existing plant used deeper understanding of the individ- to 98% of calculated capacity for
two columns in series to remove ual equipment capacities stimulated most of the time and backing off

CHEMICAL ENGINEERING PROGRESS • MARCH 1999


D E B OT T L E N E C K I N G

Uncover the opportunities


The more thoroughly you under-
stand the plant capacity, the more
ideas you should be able to come up
Sump Level Oscillation with to debottleneck the plant. Simu-
lation can help you to more quickly
and accurately evaluate these ideas.
Sump Level The more ideas you evaluate, the
more likely you should be to identify
Tray N Temperature a low-cost, low-impact debottleneck-
Oscillation ing plan. And, that means money.
Debottlenecking a plant, in many
Tray N instances, may involve changes in
Temperature
the reliability of continued operation
at higher rates so as to attain a high-
er effective onstream factor. In other
Flood cases, first principles may allow you
Indicator
judiciously to adjust operating con-
Flooding Indicator On ditions, such as pressure, to achieve
Full Flood maximum feed throughput without
exceeding the internal hydraulic ca-
pability of devices such as distilla-
Time
tion columns. CEP

■ Figure 6. The response of two variables reliably predicts impending flooding and,
so, can trigger a warning indicator.
D. B. LITZEN is principal of Virtual Ideality,
Rapid City, SD (Phone: (605) 355–9342;
Fax: (605) 355–0980; E-mail:Dlitzen@
when the incipient-flooding indica- without the incipient-flooding indi- compuserve.com), an engineering
tors turned on. Flooding episodes in cator relighting. If conditions have consulting firm. Before starting that
a typical year were reduced from an not changed sufficiently, the indica- firm, he spent nearly 16 years in various
process engineering and manufacturing
average of 12 to 2. In this case, two tor will alert to operator to stop the assignments for Shell Chemical Co.
variables provided a tipoff of incipi- ramp up. He specializes in building and extending
ent flooding: Traditional indicators such as tem- process models, particularly for nonideal
• liquid temperature 10 trays perature profile and pressure drop can chemical systems. He received a BS in
chemical engineering from South Dakota
above the bottom; and alert that flooding already is occur- School of Mines and Technology.
• liquid level in the sump of the ring, but cannot reliably warn of the A registered Professional Engineer in
column. approach of flood. In contrast, the dy- Texas, he is a member of AIChE.
These two variables showed very namic behavior of temperatures and
J. L. BRAVO is a separations specialist
repeatable oscillations with a fixed levels can be a clear indication of im- with Equilon Enterprises LLC
amplitude and period several min- peding flooding. (a Shell/Texaco/Star Alliance joint venture),
utes before every recorded flooding Houston (Phone: (281) 544–6721; Fax: (281)
episode (see Figure 6). The control 544–8123; E-mail: jlbravo@equilon.com).
Previously, he worked for Jaeger Products,
system was programmed to detect
these particular oscillations and
Literature Cited and was manager of the Separations
Research Program at the Univ. of Texas at
turn on the incipient-flooding indi- 1. Fair, J. R., and A. F. Seibert, “Under- Austin, where he received an MS in
cator. The operator then would back stand Distillation-Column Debottleneck- chemical engineering. He has extensive
off feed until the indicator turned ing Options,” Chem. Eng. Progress, 92 experience in the design and
(6), p. 42 (June 1996). troubleshooting of mass-transfer
itself off. After a small period of 2. Sloley, A. W., “Should You Switch to equipment, and has published more than
time, the operator would ramp up High Capacity Trays?,” Chem. Eng. 30 articles in the areas of distillation,
the feed again. Because actual flood Progress, 95 (1), p. 23 (Jan. 1999).
absorption, extraction, and stripping.
In addition, he is the senior author of a
point may change with time or 3. Horwitz, B. A., “Hardware, Software, book on separations technology, and an
variations in factors such as feed Nowhere,” Chem. Eng. Progress, 94 (9), instructor at AIChE’s Distillation in Practice
composition, pressure, or pressure p. 69 (Sept. 1998). course. A member of AIChE, he is a
drop, the operator then may be able registered Professional Engineer in Texas.
to get to the specified 98% point

CHEMICAL ENGINEERING PROGRESS • MARCH 1999

You might also like