Sei sulla pagina 1di 24

DISCLAIMER: PTAC does not warrant or make any representations or claims as

to the validity, accuracy, currency, timeliness, completeness or otherwise of the


information contained in this report , nor shall it be liable or responsible for any
claim or damage, direct, indirect, special, consequential or otherwise arising out of
the interpretation, use or reliance upon, authorized or unauthorized, of such
information.
The material and information in this report are being made available only under the
conditions set out herein. PTAC reserves rights to the intellectual property
presented in this report, which includes, but is not limited to, our copyrights,
trademarks and corporate logos. o material from this report may be copied,
reproduced, republished, uploaded, posted, transmitted or distributed in any way,
unless otherwise indicated on this report, e!cept for your own personal or internal
company use.
ANALYTICAL STUDY OF
LIQUID/VAPOUR SEPARATION EFFICIENCY
BY
Dr. W.D. Monnery
ChemPe! Pro"e## Te"hno$o%y L!&.
''( R)n"hr*&%e B)y NW
C)$%)ry+ AB
Dr. W.Y. S,r"e-
De.)r!men! o/ Chem*")$ 0 Pe!ro$e1m En%*neer*n%
Un*,er#*!y o/ C)$%)ry
C)$%)ry+ AB T2N 3N4
Se.!em5er (+ 2666
SUMMARY
The purpose of this work was to establish the separation efficiency of flare knock"out
drums and determine the e!pected entrained liquid droplet diameter that is carried over to
the flare. This was accomplished by using a field pilot plant skid at the Prime #est $ast
Crossfield gas plant. The skid consisted of gas and liquid inlets test separators and
entrained liquid collection in a filter%coalescer. The raw test data was entrained liquid
carryover amount as a function of gas velocity data.
$!perimental results provide incipient entrained liquid carryover velocities. The data show
that carryover rises sharply after the incipient carryover velocity and separation efficiency
drops below &&.&'. $!perimental results indicate that entrained liquid carryover average
droplet diameters are ()) to *)) microns for flare knock"out drums at +) to +)) psig.
Calculations show that the ma!imum stable droplet size can be very large at low velocities
and the calculated liquid droplet size distribution indicates that there can be substantial
variance in the droplet size and that the latter may not be very uniform. ,n order to verify
the estimated droplet sizes and distributions, further e!perimental work must include the
addition of online droplet size and distribution measurement equipment.
$!perimental results provide quantitative data for the relationship between horizontal and
vertical - factors and allowable velocities, which has to date been empirical and
sub.ective. These results show that the factor between horizontal and vertical - factors
and allowable velocities vary from about +.// to +.*0 as 1%2 varies from /.3 to *.3.
4odelling results based on using the e!perimental data give entrained liquid average
droplet diameters that are consistent with AP, 3(+ for flare knock"out drums 5/))"*))
microns6 as well as other open literature. To avoid carryover, flare knock"out drums
should be designed using a droplet size of /)) microns.
/
3.6 INTRODUCTION/BAC78ROUND
This study is part of the Alternative 7laring Technologies program sponsored by
$nvironment Canada, CAPP and PTAC. This study focuses on the efficiency of gravity
separation as it relates to flare knockout drum design and operation.
8ne of the critical issues in facilities process design and operation is vapour%liquid
separation. This is also an important issue for the improvement of e!isting flaring systems.
The problem for flaring systems is that with the uncertainty of design and operating
conditions, liquid carryover droplets may be of such a size and composition that they are
incompletely combusted. This results in the emission of many undesirable compounds to
the atmosphere, as has been outlined in previous studies and of the current 9overnment
and ,ndustry study aimed at mitigating emissions in flares.
There is an abundance of literature available on vapour%liquid separation and equipment
design, yet there has never been a systematic, comprehensive study to verify the accepted
design methodology. 1iquid vapour separator design is described in several engineering
and operating company guidelines, the 9P:A $ngineering 2ata ;ook and recent
publications such as :vrcek and 4onnery 5+&&/6 and 4onnery and :vrcek 5+&&<6. 8ther
publications of note are #atkins 5+&*06 and Talavera 5+&&)6.
The present design philosophy is to simply attempt to be conservative enough so that
separation equipment will work. =nfortunately, the definition of how conservative designs
are remains in question. 7urthermore, equipment that does function properly at design
rates may need to be re"rated for increased rates or at off"design operating conditions and
the above mentioned problem appears again 5how conservative>6. Although general
design methodology is well accepted, it is the sub.ectivity of some of the separation
parameters used in the models that are in question.
As such, the purpose of the research is to determine the efficiency of gravity separation.
Specifically, it is to determine the velocity at which carryover occurs and to estimate the
liquid particle size going to flare. This data can also be used to check current design
criteria and estimate liquid carryover at operating conditions.
<
2.6 T9EORY OF 8RAVITY SEPARATION AND SIMPLIFIED
COALESCENCE MODELLIN8
,n a liquid"vapour separation vessel, there are typically three stages of separation. The
first stage, primary separation, uses an inlet diverter to cause the largest droplets to
impinge by momentum and then drop by gravity. The ne!t stage is gravity separation of
smaller droplets as the gas flows through the vapour disengagement section of the
separator. The final stage is mist elimination, where the smallest droplets are coalesced on
an impingement device, such as a mist pad or vane pack, followed by gravity settling of
the larger formed droplets. ,n vessels like flare knockout drums, we are primarily
concerned with gravity separation since they typically have no coalescing internals, such as
mist pads.
7or gravity separation, the allowable velocity is determined so that the required
disengagement area can be determined. 7or a vertical vessel, performing a force balance
on the liquid droplet settling out provides the necessary relationship. #hen the net gravity
force, given by $q.+,
c L
L !
"
g
g #
$

6 5
=
5+6
;alances the drag force, given by $q. (,
c
! %
%
g
& % '
$

( (
6 ? % 5
=
5(6
the liquid droplets will settle at a constant terminal velocity, =T. $quating $qs. + and (
results in@
%
L !
T
'
% g
&


/
6 5 <
=
5/6
Aence, as long as the vapour velocity, =B, is less than =T, the liquid droplets will settle
out. $q. / can be rewritten as $q. <, in the well"known :ouders";rown form@

L
T
( &

6 5
=
5<6
where
%
!
'
% g
(
/
<
=
536
The drag coefficient can be calculated from a curve fit of 7ig 0"/ in the 9P:A $ngineering
3
2ata ;ook as follows@
(
/ ?
6 5 +) &3 . )


L !
%
)

=
5*6
where 2P is in ft 5microns /.(?)?+)
"*
6, densities are in lb%ft
/
and viscosity is in cP.
) ) ) '
%
% )0 . << % &(0 . <) 6 ln5 % ))0< . 3 + + = 506
The - factor from $q. 3 is the theoretical value for vertical gravity settling. ,t requires a
known liquid droplet diameter, 2P, and determination of the drag coefficient, C2. 7or
coalescing devices such as mist eliminators, the droplet diameter changes as coalescence
occurs and cannot be predicted with any accuracy. As such, the - factor for coalescing
devices is usually an empirical value, determined from e!periments. A well"known source
of empirical - factors for mist pads is the 9P:A $ngineering 2ata ;ook. Typical values
are given below in Table +. ,n addition, values can be obtained from vendors for their
particular coalescing devices.
Table + C 9P:A - 7actors
:eparator Type Pressure 5psig6 - 7actor 5ft%s6
Aoriz with Bert pad ).<) C ).3)
Bert%Aoriz with Aoriz Pad Atmospheric )./3
/)) ).//
*)) )./)
&)) ).(0
+3)) ).(+
Bacuum ).()
otes@
+. - D )./3 at +)) psigE subtract ).)+ for every +)) psi above +)) psig
(. 7or glycol or amine solutions, multiply above - values by ).* C ).?.
/. Typically use one"half of the above - values for appro!imate sizing of vertical
separators without mist eliminators.
<. 7or compressor suction scrubbers and e!pander inlet separators, multiply - by ).0 C
).?.
7or horizontal vessels, the forces of gravity and drag no longer oppose each other and a
simple vector analysis is not possible 5Talavera, +&&)6. Aowever, e!perience has shown
that horizontal velocities can be greater than the vertical terminal values, as shown in the
above table. As such, several literature publications apply multipliers to correct the
vertical terminal velocity or vertical - factor as shown in $q. ?@
*
*
( $ ( =
5?6
where subscript A indicates horizontal and subscript B indicates vertical. The factors are
either empirical or based on the fact that the time the liquid droplet takes to drop vertically
through the vapour flow area must be less than the time it takes to travel horizontally
between the inlet and outlet nozzles. This results in the correction factor 7 stated as
follows, $q. &@
+
* L $ % =
5&6
where 1$ is the effective horizontal length of travel of the liquid droplet and AB is the
vertical distance from the inlet to the liquid surface.
As such, there is considerable sub.ectivity in determining horizontal - factors.
'.6 MET9ODOLO8Y
The research program, as outlined originally by $nvironment Canada, CAPP and PTAC
was as follows@
+. ,dentify current liquid removal technologies and practices and develop a
standardized testing methodology for knockout systems.
(. ,dentify acceptable knockout performance@
-nockout efficiency will be the measure of performance. The definition of
FacceptableG knockout efficiency must be based upon what is attainable with the
technology under field conditions. H,s &&' efficiency attainable>I
/. 1ab testing of the current technology@
The identified liquid removal system5s6 must be tested to confirm that they will
meet proposed regulations and under what operating conditions. The effect of
such parameters as pressure, flow rate, compositions, ambient temperature, water
content and hydrocarbon liquids content on knockout efficiency must be
determined.
<. 7ield Pilot Testing@
Confirm the successfully lab tested liquid removal systems handle the rigours of
field operations and deliver the rated knockout efficiency.
3. Commercialization
0
3.1 EXPERIMENTAL APPARATUS AND PROCEDURE
To make the results as realistic as possible, all testing was done in the field at the Prime
#est 5previously Amoco6 $ast Crossfield 9as Plant, with no lab testing done. The
apparatus used for the e!periments was a pilot plant scale skid, which is shown in 7igures
+ " &. The apparatus consists of gas inlet piping, liquid pumping and in.ection, test
separators and a high efficiency filter%coalescer to collect entrained liquid from the test
separators. There are three horizontal test separators, each a +) inch nominal outside
diameter 5&.+/G,26 and lengths of (J*G, /J*G and 3J*G, used in order to study the effect of
vessel length on allowable velocity. The test separators each had e!ternal cage throttling
level control with the float custom made to work for fine control within the separator
dimensions and for the butane liquid. A manual globe valve located on the gas outlet
piping controls the skid pressure. The gas flow is monitored downstream using a
Aaliburton flow indicator and controlled by a manual globe valve on the gas inlet piping.
9as inlet temperature was that delivered to the skid from the gas plant. The liquid
in.ection pump is a KAC metering pump with a ma!imum flow of /0.3 gph at 03) psi.
$!periments proceeded as follows 5refer to 7igure +6. 7or each horizontal vessel
e!periment at a given pressure, a gas flow rate was calculated, a priori, to give the desired
velocity at the set liquid level for that particular e!periment. At the beginning of the
e!periment, the gas pressure was set using the outlet manual globe valve and flow was
ad.usted using the inlet globe valve thus providing the desired flow at the desired pressure.
8nce the gas flow stabilized, the liquid flow was started in an amount known to over"
saturate the gas. At this time, the skid was left until the flows and temperatures reached a
steady operating condition. 2uring the e!periment, the gas"liquid mi!ture flowed to the
selected test separator and the overhead vapour stream leaving the selected test separator
along with any entrained liquid flowed to the filter%coalescer vessel, where the entrained
liquid was collected in the boot. The liquid level in the boot was recorded at the beginning
and at the end of the test period, with the difference being the collected entrained liquid.
The gas leaving the filter%coalescer was metered such that this value along with the
amount of liquid collected provided a Fbucket and stopwatchG type e!periment. ,n order
to ensure that entrained liquid was not in the gas downstream of the filter%coalescer,
composition and hydrocarbon dewpoint measurements of the inlet and outlet gases were
taken. ,n addition, collected liquid was analyzed. The e!perimental data 5liquid level, gas
flow, liquid 5carryover6 collected6 was then used to determine the gas velocity,
corresponding separation efficiency and entrained liquid droplet diameter.
?
F*%1re 3 : Pro"e## F$o; S"hem)!*" For Se.)r)!or S-*&
&
HC(Liquid)-1-80
HC-2-80
HC (Liquid)-1-80
--
Separator
HC(Liquid)-1-80
Vertical
HC-2-80
HC-2-80
HC-2-80
HC-2-80
Separator B
Separator C
By Pass Valve By Pass Valve
LCV
LIC
LIC
I
LI
I
ilter !
"I PI
"I PI
"I PI
PCV
Separator #
"I PI
LIC
"I
PI
Coalescer
HC($as)-2-80
%i&i'( C)a*+er
CV
PI
P, Pu*p
"I
I
$as I'let
Liquid I'let
Ve't
,rai'
Horizontal Separator A
2-.** /0,0 & 11-1** S!S
,0P0 22.0 3Pa$ 4 .85C
c!6 .02** C0#0
Horizontal Separator B
2-.** /0,0 & 101-** S!S
,0P0 22.0 3Pa$ 4 .85C
c!6 .02** C0#0
Horizontal Separator C
2-.** /0,0 & -12** S!S
,0P0 22.0 3Pa$ 4 .85C
c!6.02** C0#0
Vertical Separator
117** /0,0 & 1822** S!S
,0P0 22.0 3Pa$ 4 .85C
c!6 .02** C0#0
Filter / Coalescer Separator
2-.** /0,0 & -12** S!S
,0P0 22.0 3Pa$ 4 .85C
c!6 .02** C0#0
VAJ Metering Pump
%a& .-08 $PH4 81-2 3Pa$
102** 9PSH#
F*%1re 2 : S-*& Fron! S*&e F*%1re ' : S-*& S*&e V*e;
Sho;*n% In$e!# )n& O1!$e!#
F*%1re 4 : In$e! M)n*/o$&/M*<*n% Ch)m5er F*%1re ( : L*=1*& In>e"!*on/Me!er*n%
+)

F*%1re ? : Se.)r)!or @AA F*%1re B : Se.)r)!or# @CA )n& @BA

F*%1re C : Ver!*")$ Se.)r)!or F*%1re D : F*$!er/Co)$e#"er
++
7or vertical separator, tests were performed using a <G nominal outside diameter 5/.?(*G
,26
by /J)G height vessel fitted with manual level control. The e!perimental procedure was
the same as for the horizontal separator case. The separator gas velocity was again
determined from the gas rate and vessel diameter. Again, the recorded e!perimental data,
the gas flow and liquid carryover, were used to determine the entrained liquid droplet
diameter.
3.2 EXPERIMENTAL FEED COMPOSITIONS
8riginally, it was proposed to study various liquids to mimic both lean and rich solution
gases, however due to pro.ect delays only one set of compositions was used. The
composition was established by adding sales gas and liquids from the de"butanizer
overhead such that the gas was over"saturated and carried free liquid. :ince the amount of
gas for each run was different to obtain a different superficial velocity in the test separators
but the liquid amount to over"saturate the gas was constant, the soluble versus free liquid
would be different for each e!periment. Aowever, the composition of the saturated gas
was only dependent on the pressure and temperature and would be the same for all
e!periments at the given conditions. These saturated gas compositions are given in Table
(. ,t can be seen that the +) psig composition representing the solution gas going to flare
is hydrocarbon liquid rich. This can be considered a worst case situation from the point of
view of a flare knock out drum.
+(
Table ( C :aturated 9as To Test :eparators
Component 4ole 7raction
+) psig +)) psig <)) psig
itrogen ).)/00 ).)<<* ).)<3<
4ethane ).030/ ).?&3? ).&++/
$thane ).)+?/ ).)(+0 ).)((+
Propane ).))33 ).))*3 ).))**
i";utane ).)(/+ ).))<+ ).))()
n";utane ).)0** ).)+/( ).))*)
i"Pentane ).)(&? ).))3+ ).))((
n"Pentane ).)(0+ ).))<* ).))(+
n"Ae!ane ).)+30 ).))(? ).))+<
n"Aeptane ).))(& ).)))3 ).)))/
n"8ctane ).)))3 ).)))+ ).)))+
n"onane ).)))+ ).)))) ).))))
n"2ecane ).)))) ).)))) ).))))
n"C++ ).)))) ).)))) ).))))
n"C+( ).)))) ).)))) ).))))
n"C+/ ).)))) ).)))) ).))))
n"C+< ).)))) ).)))) ).))))
n"C+3 ).)))) ).)))) ).))))
n"C+* ).)))) ).)))) ).))))
4ethylcyclopentane ).))+? ).)))/ ).)))+
Cyclohe!ane ).))+< ).)))/ ).)))+
4ethylcyclohe!ane ).)))? ).)))+ ).)))+
;enzene ).)))& ).)))( ).)))+
$thyl";enzene ).)))) ).)))) ).))))
+(<"Trimethylbenzene ).)))) ).)))) ).))))
Toluene ).)))3 ).)))+ ).)))+
p"Lylene ).)))+ ).)))) ).))))
o"Lylene ).)))) ).)))) ).))))
+/
3.3 DATA PROCESSING
7or a given e!periment, the e!perimental data consisted of pressure, temperature, actual
gas flow and the amount of liquid collected%carried over. 7rom the temperature and
pressure and the gas and liquid compositions, density and viscosity values were obtained
from the process simulator 5AM:M:
T4
6.
The following describes how data was processed to determine parameters@
:eparation $fficiency 5'6 D H1iquid ,n 5m
/
%d6 C 1iquid Collected 516%$!perimental Time
5hr6N (< hr%d % +))) 1%m
/
I % 1iquid ,n 5m
/
%d6 N +))
Belocity 5ft%s6 D Actual 9as 7low 5ft
/
%hr6 % 7low Area At :et 1iquid 1evel 5ft
(
6 %
/*))5hr%sec6
$!perimental - 7actor 5ft%s6 D Belocity 5ft%s6 % H51 " B6%BI
).3
2roplet diameter was determined by iterative calculations as follows@
+. $stimate droplet diameter
(. Calculate L from $q. *
/. Calculate C2 from $q. 0
<. Calculate - from $q. 3
3. ,f Calculated - D $!perimental -, doneE if not, ad.ust 2P and go to step (
ote, since the force balance equations are applicable to only the vertical case, vertical
data at incipient carryover was first used to determine the droplet diameters. To determine
droplet diameters from horizontal vessel data, the horizontal e!perimental - factor data
was related to vertical data at incipient carryover for +)) psig case, Table /.
,n order to further describe the entrained liquid droplets, a statistical analysis was done.
Assuming a normal distribution, to determine the standard deviation, the required
parameters were the ma!imum and the average liquid droplet sizes. Then using a
translated normal probability table, the standard deviation could be calculated, $q. +)@

A" ! #A) !
% %
, ,
/) . /

=
5+)6
where /./) is the translated O value corresponding to &&.&3' of the liquid droplets being
of smaller diameter than 2P,4AL$: and is the standard deviation. The average droplet
diameter was obtained from the e!periments and the ma!imum value was estimated by the
theory shown in Kepson et al. 5+&?&6 and provided as $qs. ++, +( and +/@
+<
L
L+
L

!,!+
#A) !
&
"
-e %
%

< . )
Pe
&+ . +
* . )
* . )
+ . )
,
+

= 5++6
where Pe is the PeynoldJs umber and #e is the #eber umber 5characterizes the
ma!imum stable liquid droplet size in two phase flow6@

!,!+
% &

= Pe
5+(6
c
!,!+
g
% &
-e

(
=
5+/6
and 91$ is the entrained liquid mass flu! in lb%s ft
(
, is the liquid surface tension in lbf%ft
and other variables and units as previously defined.
4.6 RESULTS AND DISCUSSION
,t should be noted that two sets of data were taken. The first set taken in 8ctober to
ovember +&&& indicated that there was no carryover until velocities were high enough
that all liquid entering the test separator was carried over. This was not deemed correct
and was attributed to an e!tra separation effect from a gas dome on the test separator e!it
nozzle and the level control not being sensitive enough. As such, the skid was modified so
that no e!tra separation effect occurred at the outlet nozzle, liquid level control was made
more sensitive and entrained liquid measurement in the filter%coalescer was improved with
additional valving.
A summary of the raw e!perimental data along with processed results is presented in
Appendi! ,. The incipient carryover data and results are summarized below in Table /.
ote each e!periment was repeated at least three times. The raw data is available on
request.
+3
Table / C $!perimental 2ata at ,ncipient 1iquid Carryover
:eparator Press 5psig6 Belocity
5ft%s6
- 7actor 5ft%s6 2roplet 2iameter
5microns6
Aoriz 51%2@ /.) " <.(6 +) ? ).<+0 0*+
+)) ("< 5/6 ).())")./&? 5).(&06 (0?"*+3 5</)6
<)) + ).+?? (+&
Aoriz 51%2@ *.36 +)) <"3 )./&0").<&? *+3"?/)
<)) ( )./0? 3)&
Bert +) 3"* ).(0(")./+& <*)"333
+)) /"3 )./))").<&* </&"?/0
<)) ( )./0* 3)&
As would be e!pected from field measurements there is some scatter in the data but the
trends are correct. 7or e!ample, the allowable velocity decreases as pressure increases, as
e!pected from the :ouders";rown equation with a higher vapour density and the accepted
trend of a lower - factor. ;oth the horizontal and vertical data give this e!pected trend.
Aowever, the horizontal data appears to be more consistent when the resulting - factors
are e!amined because they also decrease as pressure increases, as they should, whereas the
vertical values do not. ,t should be noted that some of the actual values of the - factors in
Table / seem somewhat high as values for +) to +)) psig are e!pected to be about ).+03
for vertical separators and slightly higher, about ).() to ).(3 ft%s, for horizontal
separators. The horizontal vessel data also show the correct trend that a longer length or
higher 1%2 ratio provides a higher allowable velocity.
8verall, e!perimental incipient liquid carryover velocity data correspond to theoretically
calculated average liquid droplet sizes that range from about /)) to ?)) microns at +) to
+)) psig. These droplet sizes are not unreasonable when compared to literature. 7or
e!ample, as stated by Capps 5+&&<6, gravity separation is only efficient for droplets of
about /03 microns or larger.
,n comparing the vertical and horizontal data, we would e!pect that the vertical incipient
carryover velocities would be lower than the horizontal ones and this is the case for the +)
psig data but not for all the +)) and <)) psig data. :trictly speaking, the force balance
applies to vertical separation and so an ad.ustment should be made to the horizontal
droplet sizes. ,n order to compare data from all the separators, the +)) psig data had to
be used as it was the most complete data set. #e selected < and 3 ft%s data for the
horizontal data versus / ft%s for the vertical data. This results in a - factor ad.ustment of
+.// for the /"< 1%2 horizontal data and +.*0 for the *.3 1%2 horizontal data. This
ad.ustment along with assuming an ad.ustment of unity for the lowest typical 1%2 in
horizontal vessels, +.3, results in the following fit of the data for the F7G factor in $q. ?@
/3) . )
6 % 5 ?*<< . ) % L $ = 5+<6
+*
These ad.ustments compare to those previously stated in the literature. 7or e!ample,
#atkins 5+&*06 stated that the horizontal ad.ustment should be about +.(3 for horizontal
vessels, commonly designed with an 1%2 of /.). Ad.usting the horizontal - factors to
vertical ones using $qs. ? and +< and then applying the force balance equations gives the
entrained liquid droplet diameter results shown in Table <.
Table < C Aorizontal ,ncipient 1iquid Carryover C $quivalent Bertical 2roplet :ize
:eparator Press 5psig6 Belocity
5ft%s6
$quivalent Bertical
2roplet 2iameter
5microns6
Aoriz 51%2@ /.) " <.(6 +) ? 33/
+)) ("< (+&"<30
<)) + +0)
Aoriz 51%2@ *.36 +)) <"3 //<"</+
<)) ( (0)
The data in Table < show that the average droplet sizes at incipient carryover for
horizontal vessels at +) to +)) psig may be more like ()) to *)) microns. :trictly
speaking, applying the vertical force balance to the horizontal case over estimates the
droplet diameter. The logic for this effect, shown in Table <, is as follows. 7or a given
velocity, a smaller diameter droplet will settle in the horizontal direction compared to the
vertical direction. This is the same effect that allows a given droplet size to settle in a
higher horizontal velocity then vertical, resulting in higher horizontal - factors.

,n addition to average entrained liquid droplet diameter, the ma!imum stable droplet
diameter and standard deviation of the droplet diameters for the +) and +)) psig cases
were calculated. The calculation spreadsheet for both unad.usted and ad.usted droplet
horizontal diameters is in Appendi! ,,, with the results summarized in Table 3.
Table 3 C Aorizontal ,ncipient 1iquid Carryover C 2roplet 2istribution
:eparator Press 5psig6 Belocity
5ft%s6
Avg 2roplet
2ia 5microns6
4a! 2roplet
2ia 5microns6
:td 2ev
5microns6
Aoriz 51%2@ /.) " <.(6 +) ? 0*+%33/ +3+( ((?%(&+
+)) / </)%//? <<<? +(+?%+(<3
Aoriz 51%2@ *.36 +)) <.3 0((%/?/ (?<0 *<<%0<0
The data in Table 3 shows that there can be substantial variance in the droplet size and that
it may not be very uniform.
+0
,n addition to the droplet size and distribution, the separation efficiency was calculated as
a function of the gas velocity. The results are given below in Appendi! ,. ,t can be seen
that carryover rises sharply after incipient carryover velocity is reached and the separation
efficiency drops below &&.&'. Capps 5+&&<6 quotes typical carryover values for gravity
separation of ).+ vol', which is equivalent to about ()) lbs%44scf of the butane liquid
used in this study. Talavera 5+&&)6 quotes :ouders";rown gravity separation carryover
values of 03 to +3) =:gal%44scf. As shown in Appendi! ,, these values correspond to
&) C &3' separation efficiency.
7inally, the data can be used for modelling. Aowever, for modelling, the e!perimental data
with the correct qualitative trends and ad.usted for the scatter was used. An anchor data
point was chosen to be a horizontal - factor of ).((3 ft%s at +)) psig, based on the
e!perimental data and e!pected values of ).() C ).(3 ft%s. The - factor was taken to be
the same at +) psig and ad.usted with pressure as ).))3 ft%s for each +)) psi pressure
change above +)) psig, based on the 9P:A rule of taking +%( of the values for mist
eliminators. Calculations are given in Appendi! ,,, and are summarized in Table * for an
1%2 D / horizontal vessel.
Table * C 4odelling 2roplet :izes
Press 5psig6 - 7actor 5ft%s6 CalcJd 2p
5microns6
CalcJd Belocity 5ft%s6 $!p Belocity 5+6
5ft%s6
+) ).((3 //( 3.( ?
+)) ).((3 (3/ (.< ("<
<)) ).(+) +&/ +.+ +
0)) ).+&3 +** ).?
+))) ).+?) +<0 ).*
otes@
+. $!perimental incipient carryover velocity.
The calculated liquid droplet diameters are based on calculating the - factor to match the
recommended values described above, with the horizontal factor applied as per $qs. ? and
+<. The droplet diameters in Table 3 at +) to +)) psig compare relatively well with AP,
3(+ recommended values for flare knock"out drums 5/)) to *)) microns6. ,n addition,
the values at higher pressures compare well to values recommended by Arnold and :ikes
5+&?*6, who analyzed industrial fabricators guidelines based on field e!perience and
recommended +<) to +3) microns for liquid"vapour separators.
The values in Table * are somewhat lower than the recommended value of 0)) microns for
a carryover of 0 lb%44scf from an unpublished study by an engineering company
54onnery, +&&36.
+?
,t should be noted that although some e!periments were run with mist eliminator pads,
because they are not used in flare knock"out drums, these results are not discussed in
detail. Aowever, the data show that for the same velocities, mist eliminator pads
decreased the entrained liquid carryover by () C //' of the values for gravity separation
alone for the test separators with an 1%2 of / C <. 7or the test separator with an 1%2 of
*.3, mist eliminator pads decreased the carryover by // C 3)' of the values for gravity
separation alone.
(.6 CONCLUSIONS AND RECOMMENDATIONS
5.1 Conclusions
+. Although there is some scatter in the data, e!perimental data verify the accepted
qualitative trends of allowable velocity decreasing as pressure increases and larger
allowable velocity for longer 5higher 1%26 horizontal vessels.
(. $!perimental results indicate that entrained liquid carryover average droplet diameters
are in the range of ()) to *)) microns for flare knock"out drums at +) to +)) psig.
/. Calculations suggest that the ma!imum stable droplet size can be very large at low
velocities.
<. The droplet size distribution indicates that there can be substantial variance in the
droplet size and that it may not be very uniform.
3. $!perimental results provide quantitative data on the relationship between horizontal
and vertical - factors and allowable velocities, which has to date been empirical and
sub.ective. These results show that the factor between horizontal and vertical -
factors and allowable velocities vary from about +.// to +.*0 as 1%2 varies from /.3 to
*.3.
*. $!perimental results show that carryover rises sharply after the incipient carryover
velocity and separation efficiency drops below &&.&'.
0. 4odelling results based on using the e!perimental data give entrained liquid average
droplet diameters that are consistent with literature.
+&
5.2 Recoen!"#ions
+. To avoid carryover, flare knock out drums should be designed using a droplet size of
/)) microns. This is an allowable vapour velocity below the ,ncipient Carryover
velocity determined in this study.
(. A continuous online droplet size and distribution measurement system must be
installed before any further e!perimental data is collected.
/. Testing of other separation and coalescing devices should be undertaken.
?.6 Nomen"$)!1re
C2, 2rag Coefficient
2, 2iameter, in or ft
2P, 2roplet 2iameter, microns
2Pipe, Pipe 2iameter, in
7, 7actor in $q. ?
72, 2rag 7orce,
79, 9ravity 7orce
g, 9ravity Acceleration, ft%s
(
gc, 2imension Proportionality Constant, 5lbf%lbm65ft%s
(
6
91$, $ntrained 1iquid 4ass 7lu!, lb%s"ft
(
AB, Bertical Aeight, ft
-, - 7actor 5$q. 36, ft%s
-A, Aorizontal - 7actor, ft%s
-B, Bertical - 7actor, ft%s
1, 1ength, ft
1$, $ffective 1ength, ft
4P, 2roplet 4ass, lb
Pe, Peynolds umber
=T, Terminal Belocity, ft%s
=B, Bertical Belocity, ft%s
#e, #eber umber
L, Parameter 2efined by $q. *
O, Translation Bariable in ormal 2istribution
, Biscosity, cP
, Pi umber
1, 1iquid 2ensity, lb%ft
/
B, Bapor 2ensity, lb%ft
/
, :urface Tension, dyne%cm
, :tandard 2eviation 5$q. +)6
()
B.6 REFERENCES
Arnold, -.$. and C.T. :ikes, F2roplet settling theory key to understanding separator"
sizing correlationsG, 8il Q 9as K., Kuly (+, +&?*, p. *).
Capps, P.#., FProperly :pecify #ire"4esh 4ist $liminatorsG, Chem $ng Prog, 2ecember
+&&<, p. <&.
9as Processors :uppliers Association, F$ngineering 2ata ;ookG, +)
th
$dition, Bol. +,
Ch. 0 5+&?06, Tulsa, 8klahoma.
AM:M:
T4
C A$A%Ayprotech 1td., Calgary, Alberta
Kepson, 2.4., ;.K. Azzopardi and P.;. #halley, FThe $ffect of 9as Properties on 2rops in
Annular 7lowG, ,nt. K. 4ultiphase 7low, Bol. +3. o. /, +&?&, p. /(0.
4onnery, #.2. 5+&&36, private communication.
4onnery, #.2. and #.M. :vrcek, F:uccessfully :pecify Three"Phase :eparatorsG, Chem
$ng Prog, :eptember, +&&<, p. (&.
:vrcek, #.M. and #.2. 4onnery, F2esign Two"Phase :eparators #ithin the Pight
1imitsG, Chem $ng Prog, 8ctober +&&/, p. 3/.
Talavera, P.9., F:electing 9as"1iquid :eparatorsG, Aydrocarbon Processing, Kune +&&), p.
?+.
#atkins, P.., F:izing :eparators and AccumulatorsG, Aydrocarbon Processing,
ovember, +&*0, p. (3/.
(+
A..en&*< I
E<.er*men!)$ D)!) S1mm)r*e#
A..en&*< II
Dro.$e! D*#!r*51!*on C)$"1$)!*on#
A..en&*< III
Mo&e$$*n% C)$"1$)!*on#

Potrebbero piacerti anche