R. S. Mohan

s. Wang

0. Shoham

Projects Separation Technology,Keplinger Hall,

The University of Tulsa,600 South College Avenue,

Tulsa, OK 74104-3189

The performance of gas-liquid cylindrical cyclone ( GLCC ) separators can be im-proved by reducing or eliminating liquid car1)'over into the gas stream or gas carryun-der through the liquid stream, utilizing a suitable liquid level control. In this study,a new passive control system has been developed for the GLCC, in which the controlis achieved by utilizing only the liquidjlow energy. A passive control system is highlydesirable for remote, unmanned locations operated with no external po~,,'er source.Salientfeatures of this design are presented here. Detailed e.tperimental and modelingstudies have been conducted to evaluate the improvement in the GLCC operationalenvelope for liquid carryover with the passive control system. The results demonstratethat a passive control system is feasible for operation in normal slug jlo~,,' conditions.The advantage of a dual inlet configuration of the GLCC is quantified for comparativeevaluation of the passive control system. The results of this study could form thebasis for future development of active control systems using a classical control

approach.

G. E. Kouba

Chevron Petroleum Technology Company,

1300 Beach Boulevard,

La Habra, CA 90631

Introduction

For many years, the Petroleum Industry has relied mainly onconventional vessel-type separators. They are bulky, heavy, andexpensive in capital, installation, and operation. Due to eco-nomic and operational pressures, the petroleum industry hasrecently shown interest in the development of innovative alter-natives to conventional separators. One such alternative is thegas-Iiquid cylindrical cyclone (GLCC). Unlike conventionalvessel-type separators, the GLCC is simple. compact, low-weight, low-cost, requires little maintenance, and is easy toinstall and operate. It is therefore gaining popularity as an easy-to-operate, economically attractive alternative to the conven-tional separator.

The development ranking of the various separation technol-ogy alternatives is shown schematically in Fig. I ( Kouba etal., 1995) .As shown in this figure, conventional vessel-typeseparators have reached their maturity, except for some minorimprovements that are being incorporated, such as new develop-ments of internal devices and control systems. Large-diametervertical cyclones and hydro-cyclones have also been used bythe industry for some time. However, recent trends in develop-ment are focused towards new types of compact separators,such as the GLCC. The compact dimensions, smaller footprint,and lower weight of the GLCC offer a potential for cost savingsto the industry , especially in offshore applications. Also, theGLCC reduces the inventory of hydrocarbons significantly,which is critical for environmental and safety considerations.

A schematic of the GLCC separator is shown in Fig. 2. It isa vertically installed pipe mounted with a downward inclinedtangential inlet, with outlets provided at the top and bottom ofthe pipe. It has neither moving parts nor internal devices. Due tothe tangential inlet, the flow forms a swirling motion, producingcentrifugal forces. The two phases of the incoming mixture are

separated due to centrifugal and gravity forces. The liquid isforced radially towards the walls of the cylinder and is collectedfrom the bottom, while the gas moves to the center of thecyclone and is taken out from the top. Currently, the GLCCfinds potential application as a gas knockout system, upstreamof production equipment. Through the control of gas liquidratio ( GLR ) , it enhances the performance of multiphase meters,multiphase flow pumps, and desanders. Other applications areportable well testing equipment, flare gas scrubbers, and slugcatchers. The GLCC is also being considered for downholeseparation, primary surface separation ( onshore and offshore) ,and subsea separation.

The strength and weakness of the GLCC are its compactness.Its strength is primarily due to its compact dimensions wherecentrifugal forces are used for separation. However, its inherentdisadvantages are that it does not offer large residence time andit cannot tolerate large surges in flow conditions. Previous stud-ies of GLCCs have been carried out for loop configurations,characterized by recombination of the gas and liquid streams atthe outlet. The significance of this configuration lies in the factthat it is self-regulating for small flow rate variations. However,for large flow variations, there is an increasing need for liquidlevel control to improve the GLCC loop operation so as toprevent liquid carryover or gas carryunder. Also, for field appli-cations other than metering, separate outlet gas and liquidstreams are needed for the GLCC. This configuration must haveliquid level control for efficient operation.

GLCC control philosophy should be aimed at developingsuitable control strategies appropriate for field applications. Thedifferent strategies which could be adopted in the field are pas-sive control, active control using classical control theory, androbust control using modern control theory. Passive control isthe simplest form of control, which does not need any externalpower source, is easy to use and is cheap. Passive control isideal for offshore applications and remote oil fields where elec-tric power is very scarce or expensive. Active control developedusing classical control theory , even though more expensive, isrelatively accurate and reliable for field operation. Modern con-trol techniques such as fuzzy logic control and neural network~

Contributed by the Petroleum Division and presented at the Nineteenth AnnualEnergy-Sources Technology Conference and Exhibition. Houston. Texas. Febru-ary 2-6. 1998. of THE AMERICAN SOCIETY OF MECHA:;ICAL ENGINEERS. Manu-script received by the Petroleum Division. October 17.1997: revised manuscriptreceived December 12.1997. Guest Associate Technical Editors: A. Wojtanowiczand C. Sarica.

Journal of Energy Resources Technology Copyright ~ 1998 by ASME MARCH 1998, Vol. 120 I 49

GAS

-

..=4j

sQ.~

~>4j

~

UQUID

--

Time

Fig. 1 S-curve for developmental ranking of separation technology

n

Fig. 2 Gas-Iiquid cylindrical cyclone configuration

control could be adopted for GLCCs which need more robust,fast-acting, and predictable controllers. In this investigation. itis proposed to design, develop and evaluate a GLCC passivecontrol system which is capable of controlling the GLCC liquidlevel, eliminating liquid carryover and gas carryunder. The spe-cific objectives' of this investigation are given in the following :

Even though several investigators have realized that the per-formance of compact separators could be improved by incorpo-rating suitable control systems, only a few control studies havebeen conducted. Kolpak ( 1994) developed a hydrostatic modelfor passive control of compact separators in a metering loopconfiguration. This model provides the sensitivity of the liquidlevel to the gas and liquid inflow rates. Genceli et al. ( 1988 )developed a dynamic model for a slug catcher. Roy et al. ( 1995 )discussed the control algorithms in digital controllers and Gali-chet et al. ( 1994) presented the development of a fuzzy logiccontroller for liquid level control. Both of the cases dealt withonly single-phase liquid flow.

From the foregoing, it could be noted that compact multi-phase separation technology research remains a critical issue forthe petroleum industry .The performance of compact separatorscould be enhanced considerably by incorporating suitable con-trol systems. An overview of the steady-state model developedfor the GLCC is presented in the forthcoming, followed by adiscussion of the experimental results. A more detailed discus-sion of the subject is available in Wang ( 1997) and Wang etal. (1998).

.Develop a steady-state model for GLCC control of liquidlevel and pressure and perform a system sensitivity analy-sis.

.Conduct experimental investigations to establish theGLCC operational envelope for liquid carryover and todetermine the liquid level sensitivity. The experimentaldata are compared with a mechanistic model-

.Design and develop a new 3-in. inner diameter (i.d.)GLCC with a passive control system. The passive controlis carried out by means of a dual edge float with throttle/orifice assembly for controlling the liquid level in theGLCC.

.Conduct an experimental study to investigate the feasibil-ity of passive control of the liquid level in the GLCC.

A brief review of the relevant literature pertaining to thisarea is provided in the next section.

Steady-State Modeling and Sensitivity Analysis

Equilibrium Liquid Level. The GLCC geometrical param-eters and nomenclature are shown in Fig. 3. The liquid levelcan be determined for the metering loop configuration by bal-ancing the pressure drops in the gas and liquid legs betweenthe gas-liquid interface and the outlet of the GLCC.

The pressure drops in the liquid and gas legs are given, re-

spectively, by

Literature Review

A detailed review of the literature on separation technologypresented by Arpandi et al. ( 1996) reveals that very little infor-mation is available about the optimum design and performanceof the GLCC. Furthermore, existing mathematical models forcyclone separators have been limited to single-phase flow withlow concentration of a dispersed phase. No reliable models areavailable (Motta et al., 1997) for cyclones (conical or cylindri-cal) that are capable of simulating a full range of multiphaseflows entering and separating in a cyclone.

CLpLQi -PLgHPJ -P2 = ( I )gc

c Q2PI -P2 = GPG G -PGgH

gc

(2)

Nomenclature

7; = 3.141592654

Subscripts

G=gasL = liquid

max = maximums = set point value

sg = superficial gassI = superficial liquid

L = length (fi)m = no. of valves and fittingsn = no. of pipe segmentsp = pressure (lblfi2)Q = volumetric flow rate (fi3Is)

V =velocity (fils)L\ = incremental deviation4> = frictional loss coefficientp = density (lbmlfi3)

c = flow coefficient ( I /ft4)d = diameter ( ft )f = friction factor9 = acceleration due to gravity (ft/s2)

g" = units conversion constant (Ibm.ft/Ibf.s 2 )

H = liquid level relative to recombined

outlet (ft)K = fitting resistance coefficient

50 / Vol. 120, MARCH 1998 Transactions of the ASME

i,G

Fig. 3 Schematics and nomenclature of GLCC loop configuration

where the liquid and gas leg coefficients are given, respectively,

by

(3)[ " 8/;Li m 8 ]CL = L -d s 2 + L Ki Z- d 4

i=1 , 7r i=1 7r I L

(4[ .8j;L; m 8 ]CG = L -d ~ 2 + L Kj ~ d 4j=1 171" j~1 71" I G

Here n and m are the number of pipe segments and the numberof valves and fittings, respectively, Equating the pressure dropin the liquid and gas legs, as given by Eqs, ( I) and ( 2) , theliquid level relative to the recombination outlet, can be solvedas follows:

(CLPLQr- cGPGQb)~

. H= (5)g(PL- Pc)

Liquid Level Sensitivity to Inflow Rates. From Eqs. ( 1 )

and (2), solving for CL and Cc at the initial conditions (set-

point)

<PGS= CGSPGQ~S/g(PL -PG)

= (6Psgc + PGgHS)/g(PL -PG) (9)

<PLS and <PGS are the equivalent hydrostatic heads correspondingto the liquid and 'gas leg frictional losses, respectively, for theset point conditions. Substituting Eqs. (8) and (9) in Eq. (5)

yields

<PLS( CL/CLS)( QL/QLr)2

-<PGs(CG/CGs)(QG/QGs)2 = Hs + 6H (10)

Equation ( 10) shows the response of liquid level in the GLCCto changes in both gas and liquid inflow rates as well as thegas and liquid leg flow coefficients. It is clear that for constantflow coefficients (i.e., CL/CLS = I and CG/CGs= I ), there exists

a family of (QL/QLS) and (QG/QGs) pairs for which there isconstant level change (6H = constant).

The sensitivity of the liquid level to liquid flow rate can beobtained by considering the liquid and gas flow coefficients andthe gas flow rate to be constant (i.e., CL/CLS = 1, CG/CGs = Iand QG/QGS = 1) in Eq. (10). Also, «PGS + HS)/<PLS = 1 could

be considered as an identity as Hs = «PLS -<PGs). Then, the

liquid level sensitivity to liquid flow rate is given by

6H = <PLS[(QL/QLS)2 -1] (11)

From Eqs. ( 8) and ( 11) , it is clear that the lower the set pointpressure drop and/or set point liquid level across the GLCC(indicated by lower <PLS)' the less the sensitivity of the liquid

level to the liquid flow rate. This observation is very significant,especially for a GLCC characterized by recombined outlet.

Proceeding along similar lines, the liquid level sensitivity to

gas flow rate is given as

6H = <PGs[1 -(QG/QGS)2] (12)

From Eqs. (9) and ( 12), it is clear that the lower the set pointpressure drop across the GLCC (indicated by lower <PGs), the

CLS = (~Psgc + PLgHs)1 PLQh (6)

CGs = (~Psgc + PGgHs)1 PGQbs (7)

in which QGs, QLs, ~Ps, and Hs are values of gas and liquidinflow rates, pressure drop across the gas or liquid leg and liquid

level, respectively corresponding to the set point.Note that adjusting the valves changes the C values. How-

ever, changing the inflow rates does not change the C values,unless the flow regime changes from turbulent to transition orlaminar. Equation (5) may be simplified for the purpose ofexpressing level sensitivity to inflow rates by substituting con-

stants <f>LS and <f>GS as defined by

<f>LS = CLSPLQhlg(PL -PG)

= (~Psgc + PLgHs)/g(PL -PG) (8)

MARCH 1998, ValJournal of Enerav Resources Technolo~y

less the sensitivity of the liquid level to the gas flow rate. Itmay also be noted that the sensitivity of the liquid level to theliquid flow rate is higher than that of the gas flow rate.

Liquid Level Sensitivity to Flow Coefficients. Similar tothe inflow rates, the sensitivity of the liquid level to the gas andliquid leg flow coefficients as a function of their correspondingvalve openings is also an important GLCC liquid level controlparameter. For constant inflow rate conditions, QL/Ql.5 = I andQG/QGS = I. Substituting these conditions in Eq. ( 10) yields

<l>l.5(CL/Cl.5) -<l>Gs(CG/CGs) = Hs + ~H (13)

The sensitivity of the liquid level to liquid leg flow coefficientcan be obtained by considering gas valve opening remainingconstant, i.e., CG/CGs = 1. As «1>GS + Hs)/<1>l.5 = 1, the liquidlevel sensitivity to liquid leg flow coefficient is given by

~H = <l>l.5[(CL/Cl.5) -1] (14)

Equation ( 14) indicates that the change in liquid level is directlyproportional to the change in liquid leg flow coefficient. FromEqs. ( 8) and ( 14) , it may also be noted that the higher the setpoint pressure drop across the GLCC, the higher the sensitivityof the liquid level to liquid leg flow coefficient.

Similarly, the liquid level sensitivity to gas leg flow coeffi-cient is given as

!).H = <PGs[1 -(CG/CGs)] (15)

Note that the sensitivity of the liquid level to liquid leg flowcoefficient is higher than that of the gas leg flow coefficient.

Sensitivity of GLCC Pressure to Inflow Gas Rate and GasLeg Flow Coefficient. Neglecting the hydrostatic head of thegas column, ( as PG ~ I) , the pressure drop across the GLCC(!).P) can be solved from Eq. (2) as

!).p = ~ = <Pbs(CG/CGs)(QG/QGs)2 (16)

where, q,Gs is a constant and is the set point pressure drop inthe GLCC given by

CGSPGQ~s<f>Gs = ~Ps

Equation (16) shows that the higher the gas flow rate abovethe set point, the higher is the sensitivity of the GLCC pressuredrop. It may be noted that the sensitivity of GLCC pressuredrop to gas flow rate is higher for higher set point pressure.Similarly, the GLCC pressure drop is directly proportional tothe gas-leg flow coefficient.

Experimental Results

In this study, a GLCC equipped with a passive control systemwas fabricated and used for the experimental investigation. Thissection discusses the specific details of the experimental facility,experimental setup and procedure, and the experimental results.The experimental results are used to compare salient modeling

predictions.

Experimental Test Facility. The experimental two-phaseflow loop consists of a metering section to measure the single-phase gas and liquid flow rates, and a GLCC test section whereall the experimental data are acquired. A standard meteringsection was used for the experimental investigations. Detailsare given in the forthcoming.

Metering Section. The metering section is comprised of twoparallel, single-phase feeder lines to measure single-phase gasand liquid flow rates. A two-phase mixture is formed at themixing tee and is delivered to the test section. Air is used asthe gas phase. which is supplied to a gas tank by an air compres-

sor with a capacity of 250 cfm at 120 psig. The gas flow rateinto the loop is controlled by a regulating valve and meteredutilizing a combination of a mass flow meter and an orifice flowmeter.

The liquid phase is supplied from a 40-gal storage tank atatmospheric pressure and is pumped to the liquid feeder linewith a centrifugal pump. Similar to the gas phase, the liquidflow rate is controlled by a separate regulating valve and ismetered using orifice and mass flow meters. The single-phasegas and liquid streams are combined at the mixing tee. Checkvalves, located downstream of each feeder, are provided toprevent the occurrence of back-flow. The two-phase mixturedownstream of the test section is separated utilizing a conven-tional separator. The gas is vented to the atmosphere and liquidis returned to the storage tank to complete the cycle.

GLCC Test Section. A schematic of the design of the GLCCequipped with passive control is given in Fig. 4. The test facilityis divided into four parts:

the 3-in. i.d. GLCC with the dual inlet configuration, asshown on the left-hand side;

2 the passive control system, shown in the center ;3 liquid carryover trap on the gas leg; and4 the recombination section, as shown on the right-hand

side.

The dual inlet of the GLCC consists of a lower inlet and anupper inlet ( refer to Fig. 4) .The lower inlet is a 3-in-dia pipe,terminating at the GLCC with an inlet having a slot/plate con-figuration with an area of 25 percent of the inlet pipe cross-sectional area. The upper inlet is simply a reduced pipe with afull bore 1.5-in. i.d. "slot" into the GLCC. The area of thecross section of the inlet pipe is also 25 percent of that of theinlet pipe.

A passive control system is designed for the lower inlet tothrottle the flow at either of the GLCC outlets by the movementof a float controlled by the liquid level. The response of sucha passive control system could be considered to be similar tothat of a control valve with quick-response characteristics. Table1 illustrates the gas and liquid throttle valve response as theliquid level changes. The passive control system consists of afloat chamber and a float assembly. The float assembly consistsof a float, two throttles and a connecting rod between the floatand the throttles. For high liquid and low gas rates, the liquidlevel in the GLCC will increase, pushing the float upwards. Theupper throttle will engage into the upper orifice, blocking thepassage of the gas and avoiding liquid carryover. As a resultof the pressure increase, the liquid level is pushed downwards.On the other hand, for high gas and low liquid rates, the liquidlevel in the GLCC decreases. This will result in the float movingdownwards, and the lower throttle engaging with the lo.werorifice. Thus, the passage of the liquid will be blocked, avoidinggas carryunder and increasing the liquid level in the GLCC.The float assembly design includes buoyancy force and gravityforce calculations. In order to make the float assembly moveup and down with the liquid level in the float chamber, the totalweight of the float assembly should be less than the buoyancyforce applied on the float.

The GLCC is equipped with a level indicator ( sight gage )installed in parallel to the body of the separator, and a differen-tial pressure transducer, which gives a measure of the liquidlevel. The separated gas and liquid phases are metered by meansof a gas vortex-shedding meter (located in the gas leg) and amass flow meter ( in the liquid leg) .The average pressure of theGLCC is measured by an absolute pressure transducer located inthe GLCC. The temperature and density of the liquid phase arealso measured by the mass flow meter.

All output signals from the sensors, transducers and meteringdevices are terminated at a central panel, which in turn is con-nected to the computer through an A/D converter. A data acqui-

) I Vnl 1?n t.",

1 ~---II: III" .

.~ ...~!

..')" gas leg :10".:.27" .:.~ -I :~ ~:5":g":~ ~!5":g".

~3/8..

1/8'LI.

IIIJIIIIIIIIIIIIIIII

IIIIIIIIiIIIII

~C~I

26..

Recomb.

-I ination

Section

60"

cc J

.--')1"

~

~~...

6'3"l'ozzle /tan;;-;;ii;j inlet

24'

.~

~12"II

43"

3" ~

Flat plate tan2ent 00 00 o~

(top view of inlet) 5"

IIII 9"

:- 79.~

Fig. 4 Schematics of GLCC loop with passive control

operating conditions of high gas and/or high liquid flow rates.Plotting the locus of the liquid and gas flow rates at whichliquid carryover is initiated provides the operational envelopefor liquid carryover. The area below the envelope is the regionof normal operating condition. In this region, no liquid car-ryover is experienced in the separator. The region above theoperational envelope represents the flow conditions for continu-ous liquid carryover.

Single Inlet. Figure 5 illustrates the operational envelopes forliquid carryover and the corresponding liquid level in the GLCCfor single lower inlet without control. The experimental results

sition setup is built into the computer using suitable software,to acquire data from the instrumentation. This setup is capableof fixing the sampling frequency at specific rates, as desired.The sampling rate was set at 2 Hz for the flow meter and 50Hz for the differential pressure transducer. Once the steadycondition is established, an arithmetic average of data collectedfor 2 min d!lration is computed as the final value of the quantitymeasured.

A regular calibration procedure, employing a high-precisionpressure pump, has been performed on each pressure transduceron a regular schedule, to guarantee the precision of measure-ments. The temperature transducers consist of a resistive tem-perature detector (RTD) sensor, and an electronic transmittermodule calibrated with an ice bath.

Experimental Results and Model Predictions. In this sec-tion, the experimental results on the GLCC performance, includ-ing the operational envelope for liquid carryover, passive con-trol system performance, and dual inlet performance are com-pared with the model prediction.

Operational Envelope for Liquid Carryover. Liquid car-ryover is the initiation of liquid entrainment into the dischargedgas stream at the top of the GLCC. It occurs under extreme

fully openhalf openfully closed

10 15 20 25 30

Vsg(flis)

Fig. 5 Comparison of operational envelope with model prediction for

liquid carryover

H, = midway between the highest and lowest acceptable level

flHmax = largest allowable deviation in liquid level

Journal of Energy Resources Technology MARCH 1998, Vol. 120 I 53

20

IJ

c

c

c~-I

~

-.-~~ Sm,.. In..(nn ,nn,~I)

~Sm,..In.., (p...w"0..~I)-.-Du.lln..,(nn,n..~I)

,

-

~~

~~~:-;~~~GLCC Press... ! i

1!

I! I

00

I'

\'sg(I!S)

Fig. 7 GLCC pressure corresponding to operational envelope

leg valve so as to bypass the gas to the downstream. At slugflow conditions, due to its quick response characteristics, thepassive control system was exhibiting oscillatory response,causing marginal instability. The performance of the passivecontrol system could be improved further ( as a future activity)through suitable modifications of the float chamber.

Dual Inlet. A dual inlet configuration significantly improvesthe performance of the GLCC for liquid carryover, as shownin Fig. 6 by the uppermost curve. The operational envelope isalmost parallel to that for a single inlet. The operational enve-lope is improved in all the three regions, namely, churn, annular .and transition. Note that in the churn region, for 7 ft/s < VS!,< lOft/ s, the performance of the dual inlet is better thanthe performance of the passive control system. The significantadvantage of the dual inlet is the effect of pre-separation. stra-tified flow occurs in the inclined inlet pipe; the upper inlet takesthe gas flow to the top of the GLCC, while the liquid flowthrough the lower inlet. For high gas flow rates, annular flowoccurs in the inlet pipe. Both inlets take the mixture of gas andliquid. The lower inlet is rich in liquid, and the upper inlet isrich in gas, which increases the efficiency of separation forliquid carryover. Thus, the operational envelope can be im-proved with a dual inlet for all flow regimes.

GLCC Liquid Level and Pressure. Figure 6 presents a plot

of GLCC equilibrium liquid level corresponding to the opera-tional envelopes for single inlet, dual inlet, and single inlet witha passive control system. The equilibrium liquid level corre-sponding to the dual inlet shows a linear trend with increase inthe gas flow rate and is found to be marginally higher than theliquid level corresponding to the operational envelope of theGLCC with a single inlet. However, the equilibrium liquid levelof the GLCC operated with a passive control system is main-tained around the inlet. Figure 7 shows a plot of GLCC pressurescorresponding to the operational envelopes for single inlet, dualinlet, and single inlet with passive control system. The GLCCpressure is found to be higher for higher gas flow rates. TheGLCC pressure for passive control is higher than that for singleinlet without control. This is because more pressure drop occursat the orifices and the throttles in the gas and liquid legs. Com-pared to the single inlet, the dual inlet can tolerate higher liquidflow rates for the same gas flow rate for the onset of liquidcarryover. Hence, the GLCC pressure is higher for dual inletconfiguration.

are presented as scattered points. The single inlet operationalenvelope is characterized by three regions, namely churn, annu-lar, and transition. The mechanistic model predictions are shownas a solid line. In the churn region, characterized by Vsg < 10ft/s, the liquid level in the GLCC is above the inlet and themechanism of liquid carryover is by churn flow. Here, as theliquid level is above the inlet, it is easier for the liquid to beblown out by incoming gas flow. The liquid level in the annularregion, characterized by Vsg > 18 ft/s, is below the inlet andthe mechanism of liquid carryover is by droplets carried in ahigh-velocity gas stream. The liquid flow rate for the onset ofliquid carryover has a linear trend with the gas flow rate in thisregion. Between the churn and annular regions is the transitionregion, characterized by 10 ft/s < Vsg < 18 ft/s, in which theliquid level is around the inlet. The mechanism of liquid car-ryover is churn flow or annular flow. In this region, the liquidflow rate for the onset of liquid carryover is fairly constant forincrease in gas flow rate. This is because the liquid level shiftsfrom above the inlet to below the inlet as the gas flow rateincreases.

The mechanistic model used for predicting the operationalenvelope for liquid carryover is a revised version of the modelpresented by Arpandi et al. (1996). Details of the model areavailable in Movafaghian ( 1997). As illustrated in Fig. 5, theliquid level and the operational envelope predicted by the modelmatch very closely with the experimental results. The marginaldeviation of the operational envelope from the model at theannular flow region could be due to the unpredictable behaviorof the gas trap added in the liquid leg of the GLCC.

Passive Control. One can intuitively expect that, when thepassive control system is activated, liquid level will be main-tained around the inlet (lower inlet). Thus, the operational enve-lope for the onset of liquid carryover can be improved in thechurn region because of the lower liquid level compared to thatof the single inlet without control. The operational envelope forpassive control is shown in Fig. 6. The operational envelope isexpanded in the churn region. For very high liquid flow rates,the passive control system fails to work, as it blocks the liquidoutlet. At this condition, the momentum force of the liquidacting on the throttle at the liquid leg is larger than the buoyancyforce of the float, which causes the throttle to block the liquidorifice. This problem could be compensated by opening theliquid leg valve to bypass the liquid to the downstream, whichcould provide an operational point ( point A in Fig. 6) for liquidcarryover at larger liquid flow rates. The passive control systemfails to work for high gas flow rate conditions, as it blocks thegas outlet. At this condition, the momentum force of the gasacting on the throttle at the gas leg is larger than the gravityforce of the float, which causes the throttle to block the gasorifice. This problem could be compensated by opening the gas

Liquid Level Sensitivil). to Inflow Rates. For a given system.the liquid level is a function of inlet gas and liquid flow rates.By balancing the pressure drop across the gas and liquid legs.the liquid level can be maintained at the same level for a combi-nation of gas and liquid flow rate conditions. Figure 8 illustratesthe combinations of gas and liquid flow rates for three different

54 I Vol. 120, MARCH 1998 Transactions of the ASME

Q,,-0886(ft'/ljQ..,-OOS (ft'/l)

~

::.r"

02 0. 0. 01

Qo/Q"

Fig. 8 Liquid level sensitivity to inflow rates

liquid levels ( set point, 6 in. below, and 6 in. above the setpoint) .The X-axis of Fig. 8 gives the ratio of the in-situ gasflow rate to the set point gas flow rate and the Y-axis gives theratio of the in-situ liquid flow rate to the set point liquid flowrate. This figure provides a measure of the sensitivity of theliquid level to the inflow rate conditions. The solid lines corre-spond to the model predictions (see Eq. (10)) and the brokenlines show the experimental result$. Given a specific liquid flowrate ratio ( say, QLI Qu = 0.8) , the change in the gas flow rateratio which causes the liquid level to increase 6 in. above theset point is a 44-percent reduction of the initial gas flow rate.For a decrease 6 in. below the set point, an increase of 131percent of the initial flow rate is required. Given a specific gasflow rate, the liquid flow rate change can also be similarlydetermined. For lower gas flow rate conditions, the model showsgood conformance with the experimental results; whereas, forhigher gas flow rate conditions, the prediction shows some devi-ations. This is because the liquid trap in the gas leg createssignificant pressure drop, which pushes the liquid level verylow. In this case, the operational envelope is extended for highgas flow rates.

equilibrium liquid level is less sensitive to liquid flowrate in the presence of lower friction losses in the liquidleg. Thus, for GLCCs characterized by higher frictionlosses in the gas and the liquid legs, and GLCCs in whichthe gas and liquid outlets are not recombined, active con-trol systems are needed for GLCC pressure and liquidlevel control.

4 Detailed experimental data were acquired to establish theGLCC operational envelope for liquid carryover and todetermine the liquid level sensitivity. The data have beencompared with the predictions of the modified mechanis-tic model by Arpandi et al. ( 1996) .A new 3-in. GLCCequipped with a passive control system has been designedand fabricated.

5 Experimental results showed that the passive control sys-tem considered does improve the GLCC performance forliquid carryover, but worked in a restricted range of flowconditions. The passive control system could be extendedfor large liquid flow rates by bypassing the liquid usinga bypass valve.

6 The dual inlet configuration of the GLCC, characterizedby a reduced pipe slot upper inlet and a sector slot/platelower inlet, provides significant merit in terms of a wideroperational envelope for liquid carryover, compared to asingle inlet configuration or passive control configuration.

Acknowledgments

The authors wish to thank Chevron Petroleum Technology

Co. and the other member companies of the Tulsa University

Separation Technology Projects (TUSTP) for supporting this

project.

ReferencesArpandi, I., Joshi, A., Shirazi, S., Shoham, 0., and Kouba, G., 1996. ' 'Hydrody-

namics of Two-Phase Flow in Gas-Liquid Cylindrical Cyclone Separators,' , SPE

30683. SPE Journal, Dec., pp. 427-436.Gillichet, S.. Foulloy, L., Chebre. M., and Beauchene. J. P., 1994, "Fuzzy

Logic Control of a Floating Level in a Refinery Tank," Proceedings, 1994 IEEEIntemational Conference on Fuzzy Systems, pp. 1538-1542.

Genceli, H., Kuenhold. K., Shoham. 0., and Brill, J. P., 1988, "DynamicSimulation of Slug Catcher Behavior,' , presented at the SPE 63rd Annual Meeting,

Houston, TX, october 2-5.Kolpak. M. M., 1994. "Passive Level Control in Two-Phase Separator,.' inter-

nal communication, Arco Exploration and Production Technology.Kouba. G., Shoham, 0., and Shirazi, S., 1995, "Design and Performance of

Gas-Liquid Cylindrical Cyclone Separators," presented at the BHR Group 7thInternational Meeting on Multiphase Flow, Cannes, France, June 7-9.

Motta. B., R., F., Erdal, F. M., Shirazi, S., A., Shoham, 0., and Rhyne, L., D.,1997, "Simulation of Single-Phase and Two-Phase Flow in Gas-Liquid Cylindri-cal Cyclone Separators," presented at the ASME Summer Meeting, Fluid Engi-neering Division. Vancouver, B.C., Canada, June 22-26.

Movafaghian, S.. 1995. "The Effects of Geometry, Fluid Properties and Pres-sure on the Flow Hydrodynamics in Gas-Liquid Cylindrical Cyclone Separators,"M.S. thesis, The University of Tulsa. Tulsa, OK.

Roy, S. and Smith, C., 1995, Better Than Averaging u\"el Control, In Tech,

July.Wang, S.. 1997, "Control System Analysis of Gas-Liquid Cylindrical Cyclone

Separators,'. M.S. thesis, The University of Tulsa. Tulsa. OK.Wang, S., Mohan, R., Shohan. 0.. and Kouba, G., 1998, "Performance Im-

provement of Gas-Liquid Cylindrical Cyclone Separators Using Passive ControlSystem,'. presented at ASME Energy-Sources Technology Conference and Exhi-bition. Houston. Texas. February 2-6.

Conclusions

The specific conclusions derived from this investigation aregiven in the following:

A steady-state model for control of the GLCC loop con-figuration has been developed v.-hich can predict the equi-librium liquid level and GLCC pressure.

2 Detailed analysis of the system sensitivity indicates thatthe equilibrium liquid level is more sensitive to the liquidflow rate and the GLCC pressure is more sensitive to thegas flow rate. Hence, liquid level control could beachieved effectively by a control valve in the liquid outletand GLCC pressure control could be achieved by a con-trol valve in the gas outlet.

3 GLCC pressure is less sensitive to inlet gas flow ratewhen lower friction losses exist in the gas leg. Similarly,

MARCH 1998, Vol. 120 I 55Journal of Energy Resources Technology