Cover Story Part 2

DISTILLATION:
Diagnosing Instabilities
In the Column Overhead
Henry Z. Kister, Fluor

James F. Litchfield, Consultant

Understanding how the equipment works — and
keeping elementary physics in mind — are the keys
roblems that involve the inter-

P nals of a distillation tower

have received a respectable
amount of attention in the
practical engineering litera-
ture.* But one cannot say the same for
ones involving the tower overhead.
This lack is surprising, because there
to eliminating effects such as siphons and unsteady
vapor-collapse rates in reflux accumulators,
condensers and other overhead components
TOWER FLAWS TYPIFY A MORE BASIC PROBLEM
is considerable potential for flawed de-
n his preface to a 1997 book, “A Working Guide to Process Equipment” [1], author
sign and for operating difficulties in-
volving elevated equipment associ-
ated with distillation columns.
I Norman Lieberman states: “The general knowledge as to how process equipment re-
ally functions is disappearing from the process industries. This is not only my opinion,
but the general view of senior technical managers, in many large corporations.”
Consider, for instance, the following Speculating about the root cause of this trend, Lieberman points that chemical engi-
three real-life examples. Although the neers, the traditional guardians of process know-how, are spread thin over today’s huge
specifics differ widely, these instances body of knowledge. This state of affairs dilutes their understanding of process equip-
shared the following attributes: ment. Minor, seemingly unimportant design flaws slip by, later turning into major hid-
• In each case, a minor flaw kept the den bottlenecks that remain unresolved and limit production for decades. When the bot-
tlenecks become entirely intolerable, a very expensive solution is often implemented: a
equipment from achieving any-
sledge hammer is brought in to crack the nut.
where near its full potential
In one case involving distillation [2], a flashing feed entering a downcomer bottle-
• A key to the correct diagnosis con- necked an entire olefins plant for 17 years. This bottleneck survived three failed fix at-
sisted of knowing the ultimate tempts. The fourth, “sledge hammer,” attempt would have replaced all the column trays
equipment capacity and recognizing by packing. A last-minute investigation correctly diagnosed the problem and imple-
that the equipment performance mented a successful and inexpensive solution.
was well short of its capability The main text of this article presents three additional instances where minor flaws led
• Another key to correct diagnosis lay to major bottlenecks, involving distillation-tower overhead systems. o
in performing a hydraulic analysis
that combined basic calculations During initial operation, the column time, the valve was found to be clean
with field tests and observations reflux flowrate was very erratic. The and in good condition.
• In each case, the correct diagnosis operators dampened this behavior by Then the fabrication drawings for
led to simple, inexpensive modifica- keeping the control valve wide open the accumulator were reviewed. It
tions. Sledge hammers were not and running the reflux accumulator at was noted that at the 10%-liquid
used (see box) approximately 10% liquid level. level, the entering feed dropped about
Over time, however, the reflux 6 ft into a shallow pool of liquid at the
Case No. 1: flowrate dropped off, so the liquid bottom of the vessel (Figure 1). The
AERATION DESTABILIZES level in the accumulator rose. The nor- liquid level was only about 18 to 20 in.
REFLUX FLOW mal reflux flowrate was reestablished above the liquid outlet nozzle.
In this facility, the reflux to a chemi- by stroking the control valve two or Descending like a waterfall, the liq-
cal-plant distillation tower flowed by three times. However, this mode of op- uid entrained vapor as it penetrated
gravity from a 30-in. (inside diameter) eration destabilized the column and the shallow pool of liquid, creating
vertical accumulator through a vortex was an operating nuisance. very fine vapor bubbles. Some of these
flowmeter and a flow control valve. bubbles then became entrained in the
The level in the accumulator was not Figuring out the cause discharging liquid, and a portion of
automatically controlled. The control valve was examined sev- those in turn became trapped down-
*See, for instance, References [1] and [3]. eral times during brief outages. Each stream, at the control valve inlet.

56 CHEMICAL ENGINEERING WWW.CHE.COM SEPTEMBER 2004

FIGURE 1. A simple rerouting
of pipe solved a problem, FIGURE 2. The addition of a small siphon breaker
converting a waterfall into a drum to a seal loop eliminated recurring siphoning
modest fountain and erratic flow

These trapped bubbles limited the reaching the accumulator outlet. The from a heavy liquid phase that then
flowrate through the valve. liquid at the bottom of the accumula- went to a distillation tower and other
The reflux pipe leaving the bottom tor would degas, reverting to non-aer- downstream equipment for final-prod-
of the accumulator had a diameter of ated liquid. In this condition, the liq- uct purification. The decanter was a
2 in. This size was generous for drain- uid would easily siphon out, and the horizontal drum, 4 ft in diameter by 8
ing non-aerated liquid, but too small accumulator level would rapidly drop. ft long, that provided well in excess of
for liquid that was aerated. Draining But thereupon, the waterfall would 1 h of residence time for the phase
the latter requires rundown lines that again aerate the bottom liquid, and separation.
are sized for self-venting flow; in other the aerated liquid flow would resume. The maximum liquid level in the de-
words, flow in which liquid descends The back-and-forth switches between canter was set by the 3-in. nozzle for
while any entrapped vapor bubbles aerated liquid flow and siphoning light-phase drawoff, which was lo-
disengage upward. caused the initial erratic behavior. cated in the decanter head, 6 in. below
An excellent correlation by Simpson the top of the decanter. The heavy
[7] and Sewell [8] for self-venting flow The cure was simple phase flowed through a block valve
is Figure 4.5 in Reference [3]. Based The accumulator happened to have an and an isolation control valve, and
on that correlation, a 2-in. line can 8-in. hand-hole, located 15 in. above then upward through a seal loop. The
drain up to 7 gal/min of aerated liquid. the bottom tangent line. The problem elevation at the top of the seal loop
Because the reflux flowrate in this was solved by rerouting the 2-in. feed was an inch or two lower than the ele-
case was 12 gal/min, the balance (5 line so as to enter the drum by passing vation of the light phase draw nozzle.
gal/min) would build up in the reflux through the hand-hole cover, as A 1-in. pressure-balance line con-
accumulator, raising the liquid level. shown in Figure 1. The rerouted line nected the top of the seal loop to the
At the same time, gas trapped at the was configured in such a way that the decanter vapor space. After leaving
valve would reduce the flow area portion inside the vessel extended to the decanter, both phases flowed to
through the valve and line, lowering the drum centerline and then bent up- their respective surge tanks at grade
the reflux flow rate. Stroking the wards, discharging upwards against a level, which was about 50 ft below the
valve vented the trapped bubbles and flat horizontal deflector baffle. This decanter elevation.
siphoned the accumulating liquid out baffle redirected the incoming liquid,
of the accumulator. spreading it sideways. This arrange- Problems
Plant operating personnel recalled ment eliminated the waterfall and With the block and isolation valves
that the accumulator had initially op- aeration, and fully restored the stabil- wide open, the decanter proved to be
erated at levels much higher than ity of the reflux. susceptible to siphoning through the
10%. At these higher levels, operation seal loop, creating erratic flow in this
had been far more erratic. The expla- Case No. 2: system. In fact, the seal loop had si-
nation for that behavior is as follows: SIPHONING IN DECANTER phoned as much as 70% of the de-
At those higher liquid levels, the OUTLET PIPES canter liquid.
“waterfall” height would diminish. In this process unit, the decanter in Because of the erratic behavior, the
This diminution and the greater pool Figure 2a separated a light liquid decanter was unable to operate at its
depth would keep vapor bubbles from phase (for recycling back to a reactor) design temperature. What’s more, it

CHEMICAL ENGINEERING WWW.CHE.COM SEPTEMBER 2004 57

FIGURE 3. In this example, the harmful of the seal loop to grade exerted strong mize turbulence and short-circuiting
pressure fluctuations could have been suction at the top of that loop. With ap- in the decanter, the internal feed pipe
avoided in the first place by not ignoring
the distillation-design literature
propriate pressure balancing, enough in that vessel was increased in diame-
vapor from the top of the decanter ter from 2 to 4 in., and the feed was
did a poor job of handling liquid surges would have become entrained in the discharged against the head, as seen
from upstream heat exchangers. rundown line liquid to raise friction in in Figure 2b. And for better control of
the rundown line and make the pres- the light-phase thickness, a weir was
Initial modifications sure at the top of the seal loop the installed in the light phase just up-
In the first try at solving these prob- same as that in the vapor space of the stream of the outlet nozzle.
lems, the size of the pressure balance decanter. But if there were no pressure As a result of these modifications,
line was increased from 1 to 2 in. and balancing whatsoever, the suction at the siphoning was eliminated and the
that of the seal loop pipe from 2 to 4 in. the top of the seal loop would have decanter outlet flow was no longer er-
These changes helped, but the erratic caused the liquid flow to rapidly in- ratic. Most important, stabilization of
flow persisted. crease and siphon out the decanter. the flow allowed the decanter to oper-
Operators were then able to lessen The observation that siphoning was ate at its design temperature, which
the siphoning by closing the seal-loop taking place meant that the pressure- was 10 to 20ºF lower than the pre-mod-
block valve halfway. But operation in balance line was not fully achieving its ification temperature, leading to major
this mode was not desirable, because intended function. Increasing the line improvement in phase separation.
whenever a feed surge occurred, some size from 1 to 2 in. had helped, but had
of the heavy phase was carried over not remedied the situation completely. Case No. 3:
into the light phase. Furthermore, The throttling of the liquid valve be- POOR HOOKUP OF
closing the valve had little effect in tween the decanter and seal loop in- HOT-VAPOR BYPASS PIPES
helping the decanter to reach the de- creased the pressure difference be- This installation consisted of the over-
sign temperature. tween the two, which dampened head for a new debutanizer column sep-
surging by further increasing vapor arating C3 and C4 hydrocarbons from
Making a hydraulic balance flow to the seal loop. However, the hy- gasoline. The column overhead vapor
A hydraulic balance over the relevant draulic balance had made it apparent was totally condensed by a battery of
equipment was made. The findings that better vapor balancing and four submerged condensers (Figure 3a).
were illuminating. siphon breaking were required. The reflux drum was elevated. Noncon-
With the pressure-balance line densables, if any, from the condensers
doing its job, the static pressure PSL The full cure were vented to the drum using 1-in.
at the top of the seal loop would equal The seal loop was replaced by a 1-by- vent lines (not shown). The tower pres-
the static pressure PD in the decanter 4-ft vertical drum (Figure 2b) that sure was intended to be controlled by a
vapor space. The small, 1-to-2-in. ele- provided good siphon breaking and hot-vapor bypass, hooked up as shown
vation difference (hHP + hLP - hSL) be- pressure equalization with the de- in that figure.
tween the liquid level in the decanter canter. The siphoning was broken by But when the tower was put into ser-
and in the top of the seal loop gave, as drawing the heavy phase from a side vice, it experienced severe pressure
expected, enough driving force to over- sump, into which liquid could only fluctuations. Maintaining a constant
come the friction head losses in the enter by overflowing a chordal weir. A column pressure was impossible, which
seal loop at normal flows. Calculations 3-in. line was installed to balance the bottlenecked the tower throughput.
confirmed that the flow resistance pressures between the top of the drum
through the seal loop, including the and the decanter vapor space. Diagnosing the problem
open valves, was extremely small. Two other changes improved the For successful pressure control by a
The 50-ft elevation drop from the top decanter operation further. To mini- hot-vapor bypass, correct piping is

58 CHEMICAL ENGINEERING WWW.CHE.COM SEPTEMBER 2004

 Authors
Henry Z. Kister, a senior fel-
low and director of fractiona-
tion technology at Fluor Corp.
(Aliso Viejo, Calif.; Phone: 949-
349-4679; e-mail: Henry.Kister
@fluor.com), has over 25 years
experience in design, trou-
bleshooting, revamping, field
mandatory. Bypass vapor must enter which contained hot vapor, and the consulting, control and startup
of fractionation processes and
the vapor space of the reflux drum bottom part, which contained sub- equipment. Previously, he was
(Figure 3b), the bypass line should be cooled liquid, simply by (cautiously) Brown & Root’s staff consul-
tant on fractionation, and worked for ICI Aus-
free of pockets where liquid can accu- touching the drum. n tralia and Fractionation Research Inc. (FRI). Au-
thor of the textbooks “Distillation Design” and
mulate, and any horizontal pipe runs Edited by Nicholas P. Chopey “Distillation Operation,” plus 70 published arti-
should drain into the reflux drum. cles, he has taught the IChemE-sponsored “Practi-
cal Distillation Technology” course 260 times. A
Most important, liquid from the con- References recipient of Chemical Engineering’s 2002 Award
denser(s) must enter the reflux drum for Personal Achievement in Chemical Engineer-
1. Lieberman, N. P., and Lieberman, E. T., “A ing, he is also a member of that magazine’s Edito-
well below the liquid surface. The bot- Working Guide to Process Equipment,” Mc- rial Advisory Board. He holds B.E. and M.E. de-
Graw-Hill, New York, 1997. grees from the University of NSW in Australia. A
tom of the drum is the most suitable Fellow of IChemE and a member of AIChE, he
2. Kister, H. Z., Hower, T. C., Freitas, P. R. de
location, but instead, extending the M., and Nery Souza Neto, J. “Problems and serves on the FRI Technical Advisory and Design
Solutions in Demethanizers and C2 Splitters Practices Committees.
liquid line downward to near the bot- with Interreboilers,” 8th Annual Ethylene James F. Litchfield, a con-
tom of the drum (Figure 3b) is an ac- Producers Conference, New Orleans, La., sultant residing in Ventura,
February 25–29, 1966. Calif. (email: litchja@-
ceptable alternative. These principles aol.com), has 37 years of ex-
3. Kister, H.Z., “Distillation Operation,” Mc-
and recommendations were first pub- Graw-Hill, 1990. perience. He was with CF-
lished almost fifty years ago [4, 5]. Braun/Brown & Root Braun
4. Whistler, A.M., Locate Condensers at for 34 years, including 15 in
Since then, they have been strongly Ground Level, Pet. Ref. 33 (3) , p. 173, 1954. research and the last two as a
staff consultant. For over a
endorsed by key recent sources ad- 5. Hollander, L., Pressure Control of Light-Ends decade, he was the firm’s
Fractionators, ISA J. 4 (5), p. 185, 1957.
dressing methods for distillation-col- technical representative to
6. Chin, T. G., Guide to Distillation Pressure the Particulate Solids Re-
umn pressure control [3, 6]. Control Methods, Hydrocarbon Proc., 58 search Institute. His expertise is in chemical en-
The initial piping design for the (10), p. 145, 1979. gineering design and multiphase flow; he has op-
timized or upgraded numerous phase separators
debutanizer (Figure 3a) defies those 7. Simpson, L.L., Sizing Piping for Process (for two or three phases), inlets to vacuum and
Plants, Chem. Eng., p. 192, June 17, 1968. atmospheric columns, and lines for two- or
principles, and violates the recom- three-phase flow. He holds a B.S. from the Uni-
8. Sewell, A., Practical Aspects of Distillation
mended practices for hot-vapor bypass Column Design, Chem. Engineer, 299/300, p. versity of California at Berkeley, and an M.S.
442, 1975. from the University of Idaho, both in chemical
piping. Subcooled liquid mixes with engineering.
vapor at its dewpoint; the vapor col-
lapses at the point of mixing; the rate of
vapor collapse varies with changes in
subcooling, overhead temperature, and
condensation rate. It is the variations
in this collapse rate that induce the
pressure fluctuations experienced, and
the consequent control-valve hunting.
Problems similar to this were re-
peatedly described in the early litera-
ture mentioned above [4, 5]. But
strangely enough, the incorrect
hookup that is shown in Figure 3a
keeps reappearing in modern distilla-
tion designs.

Switching to the sound design

To remedy the situation at the debu-
tanizer, the liquid and vapor lines
were separated. The vapor line was
modified so that it introduced the
vapor into the top of the reflux drum.
As for the liquid line, it was extended
downward, terminating a few inches
above the bottom of the reflux drum.
Figure 3b shows the modified system.
These changes fully solved the prob-
lem: the tower pressure no longer fluc-
tuated. What’s more, one could feel
the differences in temperature be-
tween the top part of the reflux drum,

An earlier version of this article was presented

at the 2003 annual meeting of AIChE, in San
Francisco last November.)
adlinks.che.com/3646-37
CHEMICAL ENGINEERING WWW.CHE.COM SEPTEMBER 2004 59