Cover Story

An Investigation of
Premature Flooding in
a Distillation Column
Surprising observations regarding flooding in the upper fractionation trays of an atmospheri
crude-petroleum distillation column are investigated
n industial distillation pro­ Henry Z. Kister
cesses, flooding is an ab­ and
normal, but relatively com­ Matthew Olsson
TPA return Fluor Corp.
mon, process condition in I sT r
which liquid accumulates in the 2 spaces ir
•--t ?—t
B II
column. The accumulation may i r "h *
IN BRIEF
be caused by excessive up­
ward vapor flow, which results "1
ll|ll

in massive entrainment, or by a
restriction in the downcomers, r
51 ±- 1.
TPA draw CRUDE TOWER
FLOODING

impeding liquid downflow in the I

in: 1
TOWER HISTORY
10 spaces
column. Flooding usually results @ 610 mm L TRAY INSPECTION
*ii 1
in dramatically reduced separa­ I I tL
1 VAPOR FLOWRATE
tion efficiency, excessive pres­ BASICS
sure drops and sometimes insta­ : •< Kero stripper vapor
Kero liquid to stripper ANALYZING TOWER
bility. In many cases, the situation ■H> !*■><» KPA return
PRESSURE-DROP DATA
leading to flooding is compli­ Ill *IS
3 spaces
cated. Here, the authors present @914 mm t: ir in REVIEW OF GAMMA
findings from an investigation of
flooding in a distillation column at
a J
U u i!» SCANS
MULTIPASS
a petroleum refinery that revealed
some surprising observations in­
- - UP=s= KPA draw MALDISTRIBUTION
MODEL
volving the loss of valve floats FIGURE 1. The diagram shows the internal geometry of the top section of the
crude-petroleum distillation column CASES MODELED
from the column trays.
trays in the fractionation section. Upon shut­EXPLANATION OF THE
Crude tower flooding down and inspection, corrosion and damage TOWER FLOOD
The investigation centered on an atmo­ were observed in the TPA trays, as expected. RELATED EXPERIENCES
spheric crude tower that experienced severeThe surprise was that the upper trays in the
corrosion in the top pumparound (TPA) and fractionation section were found clean, but CONCLUDING REMARKS
upper fractionation trays. The likely cause ofmissing most of the valve floats. The miss­
the corrosion was the entry of small quan­ ing valve floats were not found, and were
tities of water, which hydrolyzed chlorides, probably washed away or corroded away.
forming HCI that caused extensive corrosionThe absent valve floats increased the open
and damage to the 410 stainless-steel trays.area on the trays, and should have moved
Pressure-drop measurements on the col­ the trays further away from flood, but flood­
umn showed flooding taking place both ing was indeed observed. The flooding only
in the four-pass TPA trays and also in the occurred near the end of that run, and did
two-pass fractionation trays below. After not occur in the earlier runs, when the valve
some time, the pressure drop decreased. A floats of these trays were found in place.
gamma scan after the pressure drop went This raised the question: could losing the
down confirmed flooding in the upper three valve floats have promoted flood in the frac-

CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM JANUARY 2019 29

 Tower history
The upper section in the tower (Figure 1)
contains three TPA trays (Nos. 1 to 3), a
chimney tray collector, and eleven naphtha-
kerosene (N-K) fractionation trays (Nos. 4
to 14). The TPA is drawn from the chim­
ney tray, cooled, and returned to tray 1.
The remaining liquid from the chimney tray

-
•.:.v t
« L.
overflows through downcomers as reflux
to the naphtha-kerosene fractionation sec­
tion. Kerosene is drawn from a sump under
4
t* tray 14 as a total drawoff and flows to the
W kerosene stripper. Vapor returning from
the stripper enters the atmospheric tower
P J*'.

above tray 14. Trays 15 to 18 are a kero­

\ sene pumparound (KPA).
rv Trays 1-14 are conventional round, un­
caged moving-valve trays fabricated out of
FIGURE 2. On tray 6 of the tionation trays? To gain a better understand­ 410 stainless steel. The valve floats of each
column, most of the floats ing of the observation, we analyzed plant tray are 50% light valves and 50% heavy
were missing
data and applied a method we developed, valves, in accordance with a common in­
known as the Fluor multipass maldistribution dustry practice [7]. The valves on trays
model (MMM), to determine whether these 15-18 (the KPA) are small-size fixed valves.
observations can be modelled. In a previousThe TPA and KPA trays are four-pass trays
article [7], we applied the same model to at 914 mm spacing, while trays 4 to 14 are
discover multiple steady-state vapor/liquid two-pass trays at 610 mm spacing.
distributions in two-pass moving-valve trays Historically, no fouling and only a limited
at turndown. amount of corrosion were observed near the
Due to the symmetry of two-pass trays, top of the tower. In the 2015 turnaround,
a perfect split of both vapor and liquid be­ some corrosion was seen on the top tray.
tween the passes is always one possibility Most of the valve caps on tray 1 and many
and a well known steady-state distribution. on trays 2 and 3 were blown off.
Our analysis shows that uneven float re­ During the next run, conditions occurred
moval on the trays can alter the vapor and beginning in April 2016 that are believed to
liquid flows through each pass, leading to have intensified the corrosion rate. Corrosion
premature flood, but that an even removal ofnear the top of an atmospheric crude-petro­
the valves from each of the passes, as ob­ leum tower is a common experience.
served in this tower, would not lead to pre­ During early autumn 2016, the tower ex­
mature conventional flood. perienced high pressure drops in the TPA
Application of criteria for vapor cross-flow and N-K fractionation sections. By winter
channeling (VCFC) showed that the most of the same year, the high pressure drops
likely cause of the flood was the reduction in both sections came down, but at the time
in dry tray pressure drop upon valve pop- the plant was experiencing difficulty achiev­
out, which induced VCFC. The VCFC led ing the desired naphtha product (according
to premature flooding and efficiency loss. to ATSM Standard D86; 95% point). Gamma
Related experiences showed that leaving scans on trays 4 through 18 taken in Janu­
tray manways unbolted can cause flooding ary 2017 showed flooding in the upper frac­
when the loads are high and the conditions tionation trays 4-6.
favor channeling. This behavior would normally indicate
Under most circumstances, the loss of some combination of corrosion, fouling
valve floats is unlikely to cause flood. This and damage. An inspection of the tower
is by far the more common case. However, in May 2017 showed surprising results:
when the loads are high and, at the same there appeared to be corrosion and corro­
time, conditions favor channeling, loss of sion damage, but little evidence of fouling.
valves can lead to flood. Likewise, unboltedAlong with the refinery owner/operator, we
manways usually only lead to efficiency losslaunched a joint study to closely understand
or instability, but not to flood. In the cases the experience. This investigation led us to
described here, the flood occurred becausean unexpected phenomenon not previously
the towers operated at high loads and underreported. This article describes our trouble­
conditions conducive to channeling. shooting and findings.
30 CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM JANUARY 2019
 CO
Tray inspection
During a maintenance shutdown in May

C>J

o
2017, the tower was opened. The following

Pressure drop, trays 4-18 (mmHjO)

r;
rg
c
observations were made:
1. Many valves floats were missing on the x January-March 2016

M
■ *

©
o
top seven trays (for example, Figure 2). • . v- • Spring and summer 2016

N)
On trays 1 through 4, about 90% of the

©
o
valve floats were missing. On trays 5 and

8S
io
X X * X * ■ Early autumn 2016

o
o
6, about 80% of the valve floats were k_3c____ X___
- Winter 2016-2017

c
missing, and on tray 7, 30% of the floats r *
X

were missing. Some of the blown-off valve * &K — Linear (spring and summer
floats found had no legs, but most of the 2016)
>

f?
♦*
floats found had legs. The panels around x

:_:
the holes were corroded. The corrosion
appeared to be uniform across the trays. 1000
It appears that the valve floats were blown 400 500 600 700 800 900 1000 1100
N-K fractionation section vapor load (m^/tir)
off due to corrosion. One of the tray panels
with missing valves is shown in Figure 2.
2. Most of the missing floats could not be through 16, were found in good shape figure 3. The plot shows
found. The inspectors checked the trays without any irregularities. the pressure drops in the N-K
fractionation and the KPA sec­
below trays 4 to 7 as well as the chimney 4.There was limited corrosion damage to the tions of the column against
trays, but could not find most of the miss­ center and side downcomers from tray 3.vapor load in the N-K section
ing floats. It appears that most of the floats5.On tray 4, a piece of tray panel (down­
were washed away or dissolved. comer seal area) was detached.
3. Few valve floats were missing on trays 8 6.There was some foulant and scale on the
through 11. Ftard rust was found on the trays, but not a large amount. There is a
tray floor between trays 8 and 11. How­ possibility that foulant could have been re­
ever, these trays, as well as trays 12 moved during shutdown, chemical decon-

TURNKEY FRACTIONAL DISTILLATION S

BATCH AND CONTINUOUS MODES
FOR OPTIMIZED PROCESSING,
FRACTIONAL DISTILLATION SYSTEMS
FROM POPE SCIENTIFIC OFFER THE WIDEST
i—-
RANGE OF CHOICES AND CAPABILITY:
LAB. PILOT OR PRODUCTION SIZES. 0.1-1000 KG/HR
STAINLESS STEEL. HASTELLOY. GLASS. OTHER MATERIALS
* RANDOM. STRUCTURED. OTHER PACKING & INTERNALS

-I MODULAR COLUMNS. 1” - 24" DIAMETER

H .
■
AMBIENT. VACUUM OR PRESSURE CAPABILITY
OPERATION VIAPLC/COMPUTER PACKAGES OR BASIC CONTROLS
TURNKEY SYSTEMS OR INDIVIDUAL KEY COMPONENTS OFFERED

!l
EXPLOSION PROOF. ASME. CE. OTHER RATINGS AVAILABLE

These proven systems are designed for critical

• C , applications including high purity specialty chemicals

9• * and solvents, pharmaceutical intermediates, flavors,

Ml
ii m JL:
ii
hj
fragrances, foods, oils, bio-based materials, and more.

POPE'S DISTILLATION EXPERTISE - SEPARATING OUR

CLIENTS FROM THEIR COMPETITION FOR OVER 50 YEA

www.popemc.com
Botch Mode System Continuous Mode System
1-262-268-9300
olution Driven.
For details visit adlinks.chemengonline.com/73849-16
 3000
the highest-loaded tray in this section
to operate at 76% of jet flood and 56%
S E I downcomer backup (clear liquid) for the
Pressure drop, trays 4-18 (mmH2 0)

C
t abnormally high vapor load of 990 m3/h,
well above the vapor load at which the
ro

. 4^ flood occurred. The FRI (Fractionation

O

Research Inc.; Stillwater, Okla.; www.fri.

■ January-March 2016
org) jet flood correlation gave the higher
ssg
_

X . .
*♦ & * x ♦ Spring and summer 2016value of 93% of jet flood for the same
a X ■Early autumn 2016 abnormally high vapor load, which also
I* ‘X X
xxX supports no flood. The pressure drop in
X
X
the N-K fractionation section came right
down by the start of the 2016-2017
o

X
winter, and at vapor flowrates below
■T-

about 900 m3/h became even lower

1000 than during the period from January to
800 900 1000 1100 1200 1300 1400 1500 March 2016.
KPA section vapor load (m3/hr)
Since the differential pressure mea­
FIGURE 4. This plot shows the pressure tamination and steam purge, but this surement of trays 4 to 18 also includes
drop in the N-K fractionation and KPA sec­ is unlikely. the KPA section, it is necessary to ad­
tions against vapor load in the KPA section
dress the possibility of the flood initiat­
Vapor flowrate basis ing there. Using the Glitsch Equation [2]
In the following sections, as well as in as recommended in Perry’s Handbook
Figures 3 and 4, the column internal [3], corrected for the small capacity en­
vapor loads are expressed in m3/h of hancement of the small valves, we cal­
standard liquid. The internal vapor loadsculate the highest-loaded tray in the KPA
were calculated from plant instrumenta­section to operate at 92% of jet flood
tion by mass balance. For instance, theand 37% downcomer backup (clear liq­
vapor load at the top of the fractionationuid) for the abnormally high vapor load
section, as well as at the bottom of the of 1,400 m3/h through this section.
TPA, equals the overhead products Figure 4 plots the pressure drop in this
flowrates plus the vapor condensed section against the KPA vapor load. The
by the TPA. However, for hydraulic andgraph shows that the KPA vapor load
flood calculations, internal vapor and at which the flood initiates is well below
liquid loads and physical properties this 1,400 m3/h level; it is probably
were obtained from the refinery's de­ about 1,250 m3/h. Also, it shows that
tailed simulation. from January to March 2016, this sec­
tion operated at vapor loads of 1,400
Analyzing pressure-drop data m3/h and even higher with no signs
The absence of apparent fouling led of flooding. With the KPA trays found
our investigation to closely examine completely clean at the turnaround, it
the pressure drop data over the last 18 can be concluded with confidence that
months of operation. Tower pressure the flood observed on trays 4 to 14 was
drop was monitored between trays 2 a premature flood initiating in the upper
and 3 in the TPA and between trays 4 fractionation section (trays 4 to 14) and
and 18, which includes the naphtha- not in the KPA.
kero (N-K) fractionation section, as well Pressure drop for the TPA section fol­
as the KPA section. lowed similar trends. Steady operation
Figure 3 shows that for trays 4 to 18 with normal pressure drops occurred
there were no signs of flood at loads asduring January to March 2016, slightly
high as 1,000 m3/h in this section prior rising over the spring and summer. In
to March 2016. Flooding appears to ini­early autumn, the pressure drops shot
tiate at a vapor load of about 750 m3/h up dramatically, at vapor loads less than
in the autumn and even the summer the previously experienced maximum.
data. Hydraulic calculations showed By the start of the 2016-2017 winter,
that with the trays intact, there should the pressure drop fell, becoming even
not be any flooding in the N-K fraction­ lower than it was from January to March
ation section (trays 4 to 14). Using the 2016. Hydraulic calculations confirmed
Glitsch Equation [2] as recommended that with the trays intact, there should
in Perry’s Handbook [3], we calculated be no flooding in this section.
32
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM JANUARY 2019
 -

r
Historically, the naphtha product D86 95%
point has run at 140-145°C. There was no
problem keeping the 95% point within this
range throughout the spring, summer and ji
|s
even early autumn of 2016, although it ap­ * er "“1 f
!|L±2
"h *
—, r !r §a
peared to move up slightly in the early au­ £ E
I II \ I
tumn of 2016. In the winter of 2016-2017,
i
I T>
2

however, the naphtha D86 95% point in­ lUj tL._

o I
creased to more than 145-155°C, indicating
deterioration in tray efficiency in the N-K frac­T E37T" T
tionation section. *5
Analysis of the differential pressure mea­ / ft T CO

\n7
surements indicates that during the early
autumn and possibly earlier, salting out
r IB r CO

and corrosion products probably caused i ijo T s£

plugging of the TPA trays and of the upper Uj i
fractionation trays below. This accounts r (i? r
for the high pressure drops observed dur­
ing the early autumn. As time progressed, n Off
14 t co JS

the corrosion products and salts were dis­

0
solved or washed away, and the plugging
disappeared, causing the pressure drop to r Hi T
si
s
come down. The increase in open area on
the TPA valve trays after the floats popped f m.
HI
r
out caused the pressure drop in the win­
ter of 2016-2017 to be less than when the nn
floats were there. In the fractionation section,_LL._LLiJj___ LL
ir CO

CVI

the same behavior occurred at lower vapor r in U5 c CD UD UD CD

loads (below about 900 m3/h), but at higher Cl CO CO

uoijeuEj papaidQ
Lf> LO

vapor loads, the pressure drops were about

the same as those observed when the floatsat the lower vapor load, with lower-density figures. This diagram
were in place. froth. Tray 5 was also flooded, again with shows a gamma scan oi the
froth accumulating on the east at the higher “a," laken
Review of the gamma scans vapor load more dense than on the west or January 17 and 18,2017
The gamma scans (Figure 5) were conducted at the lower vapor load. However, on both
on the west side of the tower on January 17,sides, the froths were less dense than that
with vapor loads in the N-K fractionation andof those of tray 4 above. Tray 6 was also
TPA section at 900 m3/h, and on the east side flooded, but here, the froth densities on the
of the tower on January 18, with the higher east and west sides were much the same
vapor loads of 990 m3/h in these sections. and lower than those of the trays above. Tray
In Figure 5, the black and pink chords were 7 was not flooded on either side, and on the
shot on the west side at vapor loads of 900 west side (at the lower vapor load), appeared
m3/h. The blue and green chords were shot to be weeping onto tray 6. The froth density
on the east side at vapor loads in the N-K was less than that of tray 5, and much the
fractionation and TPA of 990 m3/h. Since thesame as most of the trays below.
east side was shot at a higher vapor load, Trays 8 to 14 were not flooded. Their
any flooding is likely to appear more intensefroth heights appear to have been about
on the east, and any weeping is likely to be half of the tray spacing, and the froth den­
more intense on the west. sities were not high. This is similar to tray
Figure 6 is a “Kistergram” [4], a graphic 7 above. Many trays showed liquid in their
depicting gamma scan results on trays vapor spaces, but mostly only on the west
4 through 14. Tray dimensions and froth side (at the lower vapor loads). This liquid
heights are sketched to scale. The darknessis most likely weeping from the tray above,
reflects the froth density, with darker shadesbut entrainment cannot be ruled out as
signifying denser froth. the cause. Distinguishing entrainment from
Tray 4 was flooded. On the east side at the weeping in these vapor spaces is difficult,
higher vapor load of 990 m3/h, the accumu­ although the observation of more liquid on
lated froth was particularly dense. The westthe west vapor spaces (at the lower vapor
side of this tray at the lower vapor load of loads) supports weeping. At the high vapor
900 m3/h was flooded too, but as expected loads on both scan days (above 900 m3/h),

CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM JANUARY 2019 33

 FIGURE 6. This “Kistergram"
shows the tray froth heights
and froth densities, as derived
from the gamma scans
FIGURE 7. This diagram
shows the key tray dimen-
sions for the modeled column
Side A
Side
t 610 Tray
d: spacing
63 Center Center
! weir height DC
63 Center DC P12
f U
34
West East in mm, are shown in Figure 7. The open slot
area with the valve floats in place is 10% of
Flood the active area, and the hole area of the valve
holes with all the floats removed is 13.3% of
t 1 the active area (not allowing for hole corro­
8
1 sion). Liquid to the top tray, and vapor to the
id
■ 1 bottom tray, were split equally between side
t
1 r A and B.
1 Two equalities are required for a hydrauli­
12
\ cally balanced distribution for two-pass trays
14 (Figure 8):
Lo_r
1. Pressure drop. On trays with liquid flow
weeping should have been low. Interfer­ from the center to the side, the center
ence from the lattice support beams on downcomer divides the vapor space, and
the trays may have also impacted some of the pressure on either side does not need
these radiation absorptions. to be equal. For trays with liquid flow from
What can be stated with confidence is thatthe side to the center, the vapor space is
flooding took place on trays 4 to 6, which continuous, and the pressure equalizes in
were the trays with more than 80% of the the shared space (that is, every two trays).
float valves missing. The flooding was moreFor a balanced solution, the total pressure
intense at the higher vapor loads, causing drop (through both trays) on either side of
denser liquid to accumulate on the east sidethe shared downcomer must be equal.
scans. Trays 7 through 14 (which did not 2. Downcomer backup. As the center down­
lose their valve floats) appeared to be oper­comer is shared, the calculated downcomer
ating more or less normally. backup for either side must be equal.
An observation not shown on Figure 6, but All hydraulic equations are from the pub­
that can be seen in Figure 5, is that there lished literature and are the same as those
was no flooding in the KPA section. The in our earlier work [7]. For pressure-drop
trays appeared well loaded and possibly en­calculations, we used the Klein method [5]
training, but no flood. The vapor loads in thefor valve trays, and the Summers and Cai
KPA were 1,275 and 1,370 m3/h on Januarysieve-tray pressure-drop correlation [6] for
17 and 18, respectively. trays that had lost their valve floats. We used
One mystery raised by the gamma scan the Bolles and Fair method [7] for pressure
results is that trays 4-6 were flooded, but thelosses through the tray aerated liquid. For
pressure-drop data for trays 4 through 18 indowncomer backup, we used the clas­
Figure 3 did not support flooding on these sic downcomer backup equation in Perry’s
dates. The pressure drop values measured Handbook [3].
for this section during the scans were 1,600 All cases modeled used the same vapor
mm water (10 mbars per tray) on January 17 and liquid loadings, which were those for
and 2,000 mm water (13 mbars per tray) ontray 4 on January 18, 2017. Because these
January 18. loadings are high, all light and heavy valves
were open on each tray, so multiple steady-
Multipass maldistribution model state vapor and liquid distributions based on
The tower modeled here is 8.23 meters in valves opening and closing, as previously re­
diameter, and contains two-pass sieve traysported for turndown (7), were not plausible.
at 610 mm tray spacing. Tray dimensions, Thus, for each case, the MMM resulted in
only one possible allocation of vapor and liq­
Feed
distributor
a° False
downcomer
Side B
Chimney tray
uid in the tray section, dictated by the geom­
etry that was used.
8230 Tower ID
Tray 4 }
63 Side weir
Cases modeled
When describing each case, the following
acronyms are used:
height v
Trav 5 • CDC: center downcomer
63 Side DC • DCB: downcomer backup, expressed in
64 ^ay 6 clearance.
clear liquid height divided by (tray spacing
clearance \., plus weir height)
Trav 7
• JF: percent jet flood, calculated using the
TiavS
FRI correlation
• L: percent of the total internal liquid flowrate
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM JANUARY 2019
• V: percent of the internal vapor flowrate 94% on tray 5 (center-to-side flow). On tray
Case 1, all valve floats in place. This case 7, the jet food decreased from 93% to 92%.
gave the expected performance of the in­ The percent flood stays relatively the same
tact trays. As the geometry is symmetrical, despite the large increase in open area be­
there is equal distribution of vapor and liq­ cause with the valve floats missing, the tray is
uid for all trays in the section. The percent more similar to a sieve tray than a valve tray,
jet flood was 93% and 84%, respectively, and the hydraulic diameters of the holes/slots
for the center-to-side and side-to-center are much larger. The downcomer backup de­
flow trays. The clear liquid backups in the creased by 3-9% for trays 4-7. We conclude
side and center downcomers were 56% that, based on conventional hydraulic calcu­
and 55%, respectively. lations, the loss of valves would not explain
Case 2, valve floats missing per inspec­ the flooding observed on trays 4-6.
tion. In this case, 90% of valve floats were Cases 3 and 4, asymmetric valve loss. We
missing on tray 4, 80% missing on trays 5 explored whether the maldistribution result­
and 6, and 30% missing on tray 7. This situ­ing from losing valve floats asymmetrically
ation increases the open area of tray 4 fromcould result in the flood patterns observed on
10.1% to 13.1%. The open area of trays 5 trays 4-6. These cases also show the extent
and 6 are increased to 12.6% and 12.7%, to which maldistribution propagates through
and the open area of tray 7 is increased to the tray section. Here, we report the results
11.0%. In the model, the geometry is sym­ for two of these cases. More detailed analy­
metrical, so there is still equal distribution ofsis and another case can be found in Ref.
vapor and liquid for all trays in the section. 16. In all modeled cases, asymmetric valve
The change in geometry from Case 1 to losses generated flooding patterns vastly dif­
Case 2 caused flooding in the actual tower, ferent from that observed in the actual tower.
as shown in Figure 6, but the hydraulic cal­ We therefore conclude that an asymmetric
culations showed no flood. Percent jet floodvalve loss is not the root cause of the flood
marginally rose, from 81 to 84%, on trays 4 observed on trays 4-6.
and 6 (side-to-center flow) and from 93% to Our analysis precluded downcomer

Need a Hand With Your Bulk Solids?

5
YEAR!
Innovating, modeling, and designing.
We have been your bulk material solution provider
from concept to commercialization for 50 years.

' ••

t.

A
atom,1A

* 0

8 JENIKE
&JOHANSON
SCIENCE ENGINEERING DESIGN
••***Ar»
*zar •

Solids \ Powders \ Conveying \ Processing \ Storage (978)649-3300 info@jenike.com jenike.com

For details visit adlinks.chemengonline.com/73849-11

CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM JANUARY 2019 35

 FIGURE 8. Here are key hOc.F* —, ' *Wb This time, it increases the backup on side
terms used in the multipass l «t ^ | * ^ Leer A of the CDC from tray 6. To counter, this
maldistribution model (MMM)
equations downcomer sends more liquid to panel B
AP T
p A BCT
of tray 7. This change is far smaller in mag­
AC Imaginary
hdcACT
Vact partition
V BCT
^Vlc.BCT nitude, making the liquid split 45-55% to
Lb4 panels A and B, respectively. Again, there
■
is a minimal change in vapor distribution.
The liquid maldistribution continues to zig­
Af5 AP
!/ VB4
t
zag down the tower, each time at a reduced
h <Jc.A4 --------- --------------------- ^ ^dc.84
magnitude. By the time tray 10 is reached, it
I t*5
almost completely dissipates.
✓ i^-/7
h
•a!' ip 4''
dcA5 Imaginary
w
APB5
h dc.B5
A key observation is that the pressure
drop on these trays is far more sensitive to
VA5 partition V B5 changes in vapor rate than to those in liq­
Ub ; Lee uid rate, so there is a need to make large
, changes in liquid rate to make up for small
V
APH changes in vapor rate. On the other hand,
AE
the downcomer backup is far more sensitive
---^ ----- IWB6
*| Lftf-
to the liquid flowrate.
The resulting rise of the percent jet flood
iP B7
W from 93 to 103% and the downcomer
backup from 56 to 66% (clear liquid) would
h0CJt7 bocBZ
likely flood tray 5, but on panel A only.
Case 4, 90% of the valve floats are miss­
/"T TX
,<i ing on panel A of tray 4, and 80% are
missing on panel A of tray 5. The hydraulic
Yu* .... performance for this case is shown in Fig­
O o
ure 10. For this case, the reduced dry pres­
choke. The side downcomers are gener­ sure drop through panel A of trays 4 and 5
ously sized, and maldistribution could only increases the fraction of total vapor flow to
increase downcomer inlet velocity to 0.31 these panels. For Case 3, all valves were
ft/s, which is within the operable range for present on panel A of Tray 5, so the extent
this service [3], With a tighter downcomer to which the vapor favored side A due to the
design, maldistribution can lead to down­ lower pressure drop through panel A of tray
comer choke flood. 4 was limited by the greater pressure drop
Case 3, 90% of the valve floats are miss­ through panel A of tray 5. For Case 4, with
ing on panel A of tray 4, but all other valve 80% of the valves missing on panel A of Tray
floats are present. The hydraulic perfor­ 5, the vapor favors side A to a much greater
mance for this case is shown on Figure 9. The extent. With more vapor on side A, the dry
missing valve floats on panel A of tray 4 lead pressure drop through either side of tray 4
to a slight vapor maldistribution toward side isA almost equal (a significant difference from
for trays 4 and 5 and a large liquid maldistri­Case 3). The result is a much smaller liquid
bution toward side A for trays 5 and 6. imbalance, still favoring side A. Below tray 5,
The loss of valves on panel A of tray 4 both the vapor and liquid imbalance com­
largely reduces the tray pressure drop and pletely dissipate. The rise of the percent jet
the downcomer backup on the A side of flood from 93 to 109% would likely flood tray
the imaginary partition in the CDC from tray5, but again, on panel A only.
4. To equalize the backup, the CDC sends
more liquid to panel A of tray 5. Because theExplanation of the tower flood
backup is far more sensitive to the pressure- Our analysis shows that loss of valves can
drop reduction due to the valve loss than lead to flooding only if it occurs unevenly on
to changes in liquid rate, a large increase inthe trays. Ftowever, the loss of valves ob­
liquid rate is required to counter. Solution isserved in this tower has been even, similar to
achieved with a 67-33% liquid split to pan­ that modelled by Case 2. It can therefore be
els A and B of tray 5, respectively. The higher concluded that using the conventional flood
liquid load to panel A of tray 5 increases theequations, none of the cases modeled can
pressure drop of this tray, countering the explain the flooding observed on trays 4-6.
pressure drop imbalance across trays 4 and Our previous work showed that large open
5 and reducing the vapor maldistribution. areas on trays can induce vapor channel­
The maldistributed liquid continues to tray 6.ing, such as vapor cross-flow channeling [8,
36 CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM JANUARY 2019
9], or other types [4, 70] of channeling.
Such channeling generates high vapor
velocity zones, accompanied by exces­
sive entrainment and premature flood­
SIDE A
\l SIDE B

ing, and high-liquid zones with exces­ 50% L JF 88% JF 77% TRAY 4 50% L\
sive weeping. 52 .8% Vi 50 47.2% V I
JF 103% 17.3% L-t JF 81%
For sieve and valve trays, the most 67.3% ll '32-7% L TRAY 5
common type of channeling, which is
vapor cross flow channeling (VCFC),
was shown to occur when the following
oce
66y°l67.3% L

JF 89%
■• % Vl 3.7% V
JF 85% (
55
i% ’22.5% L
47.2% V
JF 76%
50 •• V I
1 CH
! F A ■ t 32.7% LW5%

JF 96%
four factors occur simultaneously: 44.8% U /55.2% L TRAY 7
49.1% VI __1,2% V 50.9% VI
1. A high fractional hole area on the
DCBtray
53% 44J%L JF 79% JF 82%
DCB

(greater than 11 % of the active area C \

TRAY 8 55.2% L\60%
50.3% Vl 7.1% l -\%1
with sieve trays, and greater than JF 94%
51.9% l\ /48.1% L
JF 91%
TRAY 9
13-14% of the active area with sharp- DCB 50.3% Vl 0.4% V- 4%/ V '
orifice valve trays)
2.A ratio of liquid flowpath length to tray
57% '51.9% L
49.9% V1
JF 81% C
55
JF 80%
50.1% VI
i
TRAY 10 48.1% IP5%
DCB

JF 92% %f-2.6% L JF 93%

spacing above 2 49.3% ll| /50.7% L TRAY 11
3
3.A weir load exceeding 50-60 m /h m D< !■; 49.9% V 1 0.1% V50.1% V D< 5
of outlet weir 56%/49.3% L JF 80% < JF 81%
TRAY 12 50.7% L\ 57%
4. Pressure below 7 barg JF 93%
50 V ' 5
1.0% L -\%l
.' JF 93%
50.3% U <49.7% L
In the N-K fractionation section of TRAY 13
50% V I 50% V 1
this tower, the ratio of flow path length DCB56 ■ '____
DC B
TRAY 14 ,49.7% L 56 :
JF 81% JF 81%
to tray spacing is extremely high, more 50% V 1 50% V
than 5.5:1. The liquid flowrate is in the
—-SEAL PAN
VCFC range, a high, 96 m3/h m of out­
let weir to the side downcomers and so the naphtha endpoint did not largely figure 9. case 3 of the modeling had
50 m3/h m of outlet weir to the cen­ rise. Once the valve floats popped out, 90% the ,loat valves missing on panel
ter downcomers, and the pressure is VCFC set in. The VCFC flood proceeds Ao,,ray4
low (about 1 barg). The combination of without a large increase in pressure
these factors, especially with the long drop, so the pressure drop decreased
flow path length, is highly conducive to as it did in the late fall of 2016. How­
the onset of VCFC. ever, this channeling induced large inlet
For as long as the valve floats were inweep that caused the tray efficiency of
place, the open slot area of the valves trays 4-6, and possibly 7, to drop sig­
was about 10% of the active area, nificantly, causing the observed rise in
making them resistant to VCFC. Hart­ naphtha endpoint.
man [7 7] observed that at the high ratio Davis [72] explained the VCFC phe­
of flow path length to tray spacing of nomenon as a force balance between
3.6, even sharp-orifice valve trays with the hydraulic gradient that tends to
about 13-14% open slot area can ex­ channel the vapor and the dry pressure
perience VCFC. With the higher ratio drop that tends to evenly distribute the
in this tower (5.5:1), it appears that the vapor. Davis proposed a rule of thumb
relatively low open area of these trays that states that to prevent the onset of
(10%) kept them well below the thresh­ VCFC, the hydraulic gradient needs to
old of VCFC, thus protecting them from be kept below 0.4 of the dry pressure
VCFC. However, once the valve floats drop. For the N-K fractionation trays,
were lost, the trays became sieve trays based on the January 18 loads, and the
with about 13% open area, the trays valve floats in place, we calculated a
entered the VCFC range. The onset of dry pressure drop of about 110 mm of
VCFC upon major valve loss was there­liquid, and hydraulic gradients of about
fore the root cause of the premature 50-60 mm. Once the valves popped
flood, along with poor efficiency ex­ out, the hydraulic gradients remained
perienced in this section. VCFC leads essentially the same, but the dry pres­
to tray inlet weep [8,9], which is highly sure drops were halved (to about 50
detrimental to tray efficiency. mm of liquid), which is about the same
Initially, when the trays experienced as the hydraulic gradients, and is in
plugging or salting out, the pressure major violation of Davis' criterion. It is
drop rose, signifying flooding, but there certain that under these conditions,
was not a great deal of efficiency loss, VCFC set in.
37
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM JANUARY 2019
 As in the crude tower, the most likely
explanation for the flood is channeling,
SIDE A
II SIDE B
CHIMNEY TRAY
in this case induced by the gap gener­
ated at the open manways rather than

f 50% L

JF 109%
58.7% vf
JF 96% C
■]=:
2.8% l•raj
JF 69%
41.3% VI
TRAY 4 50% L

JF 78%
1 the hydraulic gradient. Again, all the
conditions conducive to channeling (as
described above) existed in the tower,
52.8% l\ /47.2% L TRAY 5 except for the large open area. Once the
e
I
53% 52j8%^
58.7% V I 8.9% V

T:
JF 81%
.i
<:
155
JF 80%

%f-3.9% l
'%vi

50.2% 0 l
4
TRAY 6 47.2% l\ 9%
p manways were removed, the open area
largely increased, causing channeling to
JF 92%
48.9% ll Sl.1% L
JF 93% set in and initiate premature flood.
TRAY 7
DCB 49.8% V 1 Dt B
;
55%f48.9% L JF 80%
C
JF 81%
TRAY 8 Sl.1% U \57% Concluding remarks
55 1 The analysis reported in this article
JF 93% 1.5% L -1%7 JF 92%
50.4% L 149.6% L TRAY 9 shows that loss of valves, or open man­
(X p. 50.1% Vl 0.1% V-^. 49.9% VI ['I p ways, can lead to channeling, which in
JF 81%
56 '50.4% L C
JF 80%
TRAY 10 49.6% lW° turn, can lead to premature flooding.
: 5 50% V t
50% V t Whether the channeling will result in
JF 93% \%f-0.5% L JF 93%
49.9% l 150.1% L TRAY 11 flood depends on the operating rates
Dl B 50 v : 50 V • W B
56% JF 81% JF 81% and whether the tray geometry makes
10 TRAY 12 50.1% LP\ 6%
it prone to channeling. In the case de­
50% V 1 0.1% L -+%/
JF 93% JF 93% scribed, it can be stated with confidence
50% L I / 50% L TRAY 13
t" P
I
56% 50% L
50% V ’

50% V 1
JF 81%
50% V 1
JF 81%
50% V T
TRAY 14
DJ B
50% L 56%
\
___— SEAL PAN
FIGURE 10. Case 4 of the modeling had Related experiences
90% of the float valves missing on tray 4 A colleague [ 7 4] of the authors brought
and 80% missing on tray 5
to our attention some related experi­
ences. In several towers, there have
been experiences in which leaving the
tray manways uninstalled led to flood­
ing. He described one experience in
particular. A 3-m, inner dia. chemical
distillation tower operating at about 1
barg with 80 single-pass sieve trays at
460 mm tray spacing and weir loads
of 80-90 m3/h m of weir length was
returned to service after a routine
turnaround in which no modifications
to the trays were performed. The tray
manways were dismantled for inspec­
tion but were not re-installed following
the inspection. The tower flooded at
70% of the rated jet flood, which did
not happen prior to the turnaround.
The flood was recognized by exces­
sive entrainment from the top of the
tower, which doubled the measured
reflux flowrate for the normal amount
of reboiler steam. It was difficult to
keep the level down in the reflux drum.
Like in the crude tower, the pressure
drop was normal. Not installing the
manways can be expected to drop the
tray efficiency, as it usually does, but
increasing the open area can be ex­
pected to give the trays a larger mar­
gin from flooding.
38 CHEMICAL ENGINEERING WWW.CHEMENGONUNE.COM JANUARY 2019
that the loss of valves led to VCFC,
which in turn led to flood.
Under most circumstances, loss of
valve floats is unlikely to cause flood. This
is by far the more common experience.
However, when the loads are high and
at the same time, conditions favor chan­
neling, loss of valves can lead to flood.
Likewise, unbolted manways usually only
lead to efficiency loss or instability but not
to flood. In the cases described here, the
flood occurred because the towers oper­
ated at high loads and under conditions
conducive to channeling.
Our modeling work showed that for
two-pass trays, an even loss of valves
on both tray panels is unlikely to lead
to flood in the absence of channeling.
However, an uneven loss of valves on
one of the panels is likely to move this
panel closer to flood, and if the loads are
high enough, initiate flood on the tray.
Finally, the study discussed here dem­
onstrates that combining field data and
gamma scans with hydraulic and mal­
distribution analysis using our recom­
mended procedure [7,75) is a powerful
tool for analyzing and diagnosing col­
umn problems. This approach has been
successful for diagnosing and analyz­
ing multipass tray maldistribution, un­
expected efficiency loss at turndown,
channeling, and uneven fouling. ■
Edited by Scott Jenkins
References
1. Olsson, M., and Kister. H.Z., ‘Can We Counton Good Turn­
down in Two Pass Moving Valve Trays' Chem. Eng Piogr .
p.43, November 2018.
2. Qitsch.lnc, Bulletin 4900.6th Ed.. Dallas, Texas 1993.'Preventing Mattslributkxi in Multi Pass Trays'. Chem. &W
3. Kistef. H.Z.. R Mathas DE Stelnmeyer.W.R. Penney, Prog.,VS,April
Mon­2010, pi 32.
ica! and J. R, Fair. 'Equpment for Distillation. Gas Absorption.
16. Olssori. M„ and H.Z. Kister, "Can Loss of Valve Floats Lead
Fhase Dispersion, and Phase Separation', Set 14,toinPremature
Fteny andFlood? Paper Presented at the DisMafon
Green's 'Chemical Engineers' Handbook.' 9fri ed.. Symposium.
2018. AJCtiE Spring Meeting. Orlando. Florida, April
22-26.2018.
4. Kister, HZ.. Apply quantitative Gamma Scanning to High- Full copy is available from the author.
Capacity trays Chem Eng. Progr.. p. 45. April, 2013.
Authors
5. Klen. GF„ 'SimpMed Model Calculates Valve Tray Pressure
Drop', Chem. Eng.. May 3.1982. p 81 Henry Z. Kister is a senior fellow and
the director of fractionation technology
6. Summers D. R., and T. J. Cai, 'Dry Tray Pressure Drop of at Fluor Corp. (3 Polaris Way, Aliso Viejo.
Sieve Trays Revisited,' Chem. Eng., August 2017, p. 38.
r t Calif.; Phone: 949-349-4679; Email:
7. Bolles W. and J.R. Fair, in J.M.McKefta (Ed.) 'Encyclopedia henry.kister@tluor.com). He has over
of Chemkal Processing and Deskyn'.vol. 16. p. 86.1982. 30 years experience in design, trouble­
shooting, revamping, field consulting,
8. Kister, H. Z., K. F. Larson, and P. E. Madsen, Vapor Cross
control and startup of fractionation pro­
Row Channeling on Sieve Trays: Fact or Myth?, Chem. Eng.
Progr., p.86, November 1992. 1 cesses and equipment. Kister is the
author of three books, the distillation
9. Kister. H. Z„ Can Valve Trays Experience Vapor Cross Flow
equipment chapter in Perry's Handbook, and over 100 arti­
Channeling?. 7teGhenwaf£nginear.g18.June10.1993. cles, and has taught the IChemE-sponsored "Practical Distil­
10.Kister, H. Z, N O'Shea, and D. Cronin. Loss into lation
Gain Technology'
in course more than 500 times in 26 coun­
High-Capacity Trays Part 2: Reverse Vbpor Cross tries.flow
A recipient
Chan­ of several awards, Kister obtained his B.E.
neling, PTO, p.27.03.2016. and M.E. degrees from the University of New South Wales in
Australia. He is a Fellow of IChemE and AlChE, Member of the
11.Hartman, E. L., New Mllennium, Old Problems:NAE. Vapor andCross
serves on the FRI Technical Advisory and Design
Row Channelng on Valve Trays, in 'Distillation Practices
2001: Fron­ Committees.
tiers in a New Millennium.' Proceedings of Topical Confer­
ence; p. 108. AlChE Spring National Meeting, Houston. Matthew Olsson was a part of the
Texas, April 22-26.2001. Distillation Expertise Team at Fluor
Corp., in both Sugar Land. Tex., and
12.Davies, J. A., Bubble Trays - Design and Layout, Pet. Ret Aliso Viejo, Calif. He has now moved to
29(8). p-93,1950, and 29(9), p.121,1950. Eastman Chemical Co. in Longview,
13.Kister. H. Z.. 'Is the Hydrauic Gradient on Sieve and Valve Tex., where he is a process design en­
Trays Negligible?' Paper presented at the Topical Conference gineer. He has ten years of experience
on Distillation, AlChE Meeting. Houston. Texas, 2012. in design, troubleshooting and revamp­
14. Olsson, F.R., Private communication, March, 2018. ing fractionation processes and equip­
ment. Olsson holds a B.S.Ch.E. from
15.Kister. H. Z., R. W. Dionne. W. J. Stupin. and Texas
M. Olsson.
ASM University in College Station, Tex.

INNOVATION
TM

FLEXIPRO
technology,
the next generation
fixed valve tray.
The reliability of a large fixed valve that delivers enhanced performance
combined with an operating range close to that of a movable valve.
The FLEXIPRO™ valve tray exceeds the performance of similar fixed valve trays:
• Increased tray efficiency and higher capacity
• Higher turndown ratio with no increase in pressure drop
• Provides more flexibility at stable operating conditions without loss of tray efficiency
• Enhanced push and sweeping effect over the tray deck to remove solid deposits
- Mitigates the risk of fouling and achieves longer run lengths

YOU CAN RELY ON US. 0*80

If KOCH-GLITSCH
0m
United States 316-828-5110 | Canada 905-852-3381 | Italy 4-39-039-638-6010 | Singapore 465-6831-6500
For a complete list of our offices, visit our website.
www.koch-glitseh.com
For related trademark information visit http://www.koch-glitsch.com/trademarks. Patent pending.

For details visit adlinks.chemengonline.com/73849*13

CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM JANUARY 2019 39