Chemical Engineering November 2013 - 31

Troubleshooting " book [3], not one
came from a sour water stripper.
The theory of foaming emerged as a
front-runner, having significant support,
which was by no means conclusive.
6. An installation error. There was
not much evidence supporting that
an installation error was responsible
for flooding. The tower was wellinspected
prior to startup and no
problems were found. Nonetheless,
there were some concerns and we
therefore kept installation errors on
the troubleshooting list. Specifically,
there was a suspicion of incorrect
downcomer clearance [which could be
manually adjusted (with a tolerance
of ±2 mm)]. The inspection records
were checked and the inspection report
only indicated that the recommended
clearances had been met, but
with no values recorded.
7. Tray damage. Damage is not uncommon
in steam-water operations [8],
such as in sour water strippers. Steam
rapidly condenses when it comes into
contact with a large quantity of cold
water. The pressure in the contact
zone plunges and induces a rush of
steam from above and from below into
the contact zone. This rush imposes
a large downward-acting differential
pressure across the upper trays, and a
large upward-acting differential pressure
across the lower trays. Tray damage
may result if any of these differential
pressures exceeds the allowable
stress of the trays. Trays above the
contact zone will be deformed downwards,
and those below the contact
zone will be deformed upwards. This
damage has been previously reported
in sour water strippers [3].
The above tray damage theories
hinge on cold water entering the stripper.
In this column, the cold water
damage incident needed to have occured
during startup because the
column never reached the design feedrates.
During normal operation, the
feed to the tower was hot, 90 to 95°C.
At these feed temperatures, damage
is unlikely. At start-up, live steam
was first brought in at a slow rate,
the reflux and condenser were commissioned,
and then the tower bottom
was circulated through the preheater
back to storage. When feed eventually
entered the tower, it was hot.
Nonetheless, there were short periods
in which the preheater was bypassed,
dropping the tower feed temperature
from 90 to 40°C within a few
minutes. Stepping up the feedrate at
the lower temperature, or a sudden
drop of the feed temperature (for example,
by opening the bypass valve too
fast), or a sudden drop in steam flowrate
while the feed was cold, could have led
to tray damage. Also, there were some
" teething problems " with the stripping
steam piping and with leaking gaskets
on the plate-and-frame exchangers,
and the column went through a
number of startup/shutdown cycles, as
well as some feed-bottom exchanger
switches. Each of these could have led
to feed changes and dropping of feed
temperature over short periods, possibly
resulting in the type of damage
described above. So while there was
no reason to suspect damage, it would
have been very conceivable.
With the first three theories above
ruled out or considered highly unlikely,
four theories remained, with
foaming being a slight front-runner,
and the installation error bottoming
the list. Further work on narrowingin
on the root cause, the most likely
explanation, and the path forward are
described below.
Hydraulic analysis
Our hydraulic analysis looked at the
design case as well as the operation
bottleneck, and considered the possibility
of foaming. Our calculations
were based on methods by Fractionation
Research, Inc. (FRI; Stillwater,
Okla.; www.fri.org) and Perry's Handbook
[6]. We compared our findings to
values from the vendor rating sheets.
A. At design loads. At design loads,
the vendor calculated 76% of jet flood,
while we calculated 70 to 76% of jet
flood from the FRI valve tray and
equivalent sieve tray correlations as
well as from published methods [6].
The numbers are the same, but there
was a huge difference: the vendor's
calculation applies the 0.7 system factor
specified in the column data sheet.
Our calculations did not apply a system
factor. If we applied the specified
system factor to our calculations, the
trays would rate at 100% to 109% of
jet flood.
The FRI jet-flood correlation is
known as the best in the field, gives
predictions within 10-15% accuracy,
and concurs with published methods
with which we have had excellent experience.
Also, the FRI calculations,
as well as the published method, reflect
the industry's experience that
sieve and valve trays of equivalent
open area give much the same jet
flood capacity. We concluded that our
calculations are closer to reality than
the vendor's.
The practical implication of this discrepancy
is that the current design is
suitable for a non-foaming system, but
does not follow the industry's practice
of using a system factor of 0.5 to 0.7
for sour water strippers [6].
The downcomer backup we calculated
is 264 mm compared to the 199
mm calculated by the vendor. With a
total height of the downcomer plus
outlet weir of 660 mm, both numbers
are adequate for a non-foaming system.
The larger of the two numbers
(264 mm) is too tall for a foaming
system for which an aeration factor
of 0.4 (or even less) has been recommended
[6].
The downcomer inlet velocity is simply
the downcomer inlet clear-liquid
flowrate divided by the downcomer
inlet (top) area. Our calculation gave
a downcomer inlet velocity of 0.122
m/s. This velocity is adequate for a
non-foaming system. For a foaming
system, the maximum recommended
downcomer inlet velocity is 0.061-
0.076 m/s [6]. It can be concluded that
the downcomer design is adequate for
a non-foaming system, but grossly inadequate
for a foaming system. The
downcomer inlet areas are too small
to accommodate the liquid rate.
Our design case evaluation therefore
confirmed that the trays/downcomers
were not adequately designed
for foaming conditions. The downcomer
velocities are far too high and
the system factor application in the
design is questionable.
B. At operating loads. We also hydraulically
rated the column based
on operating data at the loads just
before flooding was observed. It is
important to recognize that this occurred
at about half the design loads.
Our calculations show that for these
ChemiCal engineering www.Che.Com november 2013 31
http://www.fri.org http://www.Che.Com
Chemical Engineering November 2013

Table of Contents for the Digital Edition of Chemical Engineering November 2013
