Académique Documents
Professionnel Documents
Culture Documents
58 JULY 1998 •
Several alternatives for GLCC liquid-level control have been
implemented. For example, a commercial multiphase measure-
ment system has used conventional control equipment successful-
ly to maintain a tight control on liquid level by controlling the gas-
outflow rate of the GLCC. Another project explored low-power
alternatives to conventional level controls that exploit hydrostatic-
head difference in the GLCC to operate the controls.8 A recent
study examined GLCC performance with a passive control system
that uses only the flow energy and no external energy.5
Crucial future work is to develop robust, active liquid-level-con-
trol strategies. Because of the smaller residence time of the compact
separator and the stringent response time requirement of the control
valve, this is not a simple extension of the control technology avail-
able for large vessel-type separators. The strategies should enable the
GLCC to handle slugging, surging, and a wide range of flow rates,
from essentially full-gas-flow to full-liquid-flow conditions.
• JULY 1998 59
vLs, ft/sec
yL0
vgs, ft/sec
vgs, ft/sec
Fig. 3—Operational envelope for liquid carry-over in a 3-in.
Fig. 2—Zero-net-liquid-flow holdup in air/water system.2 GLCC operated with air and water.2
Another GLCC combination is two GLCC’s used in series. A the- tation, ability to model an entire system, and suitability for PC
oretical work successfully predicted that a second-stage GLCC on operation. Consequently, these models are more accessible to engi-
the gas leg would increase the liquid-carry-over performance of the neers as a design tool than are CFD models.
two-stage GLCC to the onset of the mist-flow boundary, which was
anticipated as the theoretical performance limit.4 It was found that Mechanistic Modeling. The ultimate aim of modeling work to date
at very low liquid loading, the centrifugal forces on the liquid has been to predict the operating envelope for the GLCC with respect
would allow the operational envelope to push well beyond the fully to liquid carry-over in the gas stream and gas carry-under in the liquid
developed annular-mist-flow boundary. stream. Each fluid-flow path has its own particular set of calculations.
A commercial measurement system has been developed that uses The starting point for either calculation path is the global distribution
a second-stage horizontal pipe separator to remove any small gas of gas and liquid in the GLCC, namely, the equilibrium liquid level.
bubbles that may have passed along with the liquid underflow from Equilibrium Liquid Level. The equilibrium liquid level in the
the GLCC. This has allowed extension of the system operation GLCC is determined by the pressure drop between the gas and liq-
beyond the “normal” operating range of the GLCC for complete uid outlets. Because the frictional losses in the GLCC are low, the
gas/liquid separation. equilibrium liquid level is a reasonable indication of the amount of
liquid in the GLCC. The model is based on a pressure balance on
Miscellaneous Hardware Improvements. Several other potential the gas and liquid legs. Ref. 2 gives details of this model.
improvements have been considered; however, we have not dis- Vortex Shape and Location. The shape and location of the vortex
cussed them here because little or no performance information is are important for prediction of both liquid carry-over and gas carry-
available. These include a variable inlet-slot area and the config- under. The vortex model assumes rigid-body rotation (i.e., a linear
urations of the gas and liquid outlets. tangential-velocity profile in the radial direction).2 Coupling the cal-
culations for equilibrium liquid level and vortex shape makes deter-
SIMUL ATION mination of the location of the vortex and the height of the vortex
In the past, performance predictions of GLCC separators have crown possible. This model of the global distribution of gas and liq-
been carried out on the basis of experience, rules of thumb, and uid provides the groundwork for the performance models.
empirical correlations. These methods are limited in their ability Liquid Carry-Over. Liquid carry-over in the gas stream is largely
to be extrapolated to different flow conditions and untried appli- dependent on the flow pattern in the upper part of the GLCC. Flooding
cations. Currently, efforts are under way to develop mechanistic may occur in the GLCC at high liquid levels and low gas rates, produc-
models for the GLCC and conduct computational fluid dynamic ing bubbly flow. The unstable liquid oscillations, characteristic of churn
(CFD) simulations. flow at moderate gas rates, may splash liquid into the gas outlet. Liquid
Mechanistic modeling offers a practical approach to GLCC design can also be carried out in droplets at the onset of annular mist flow at
and performance prediction. Simplifying assumptions are used, but, high gas rates. At very high gas rates, the centrifugal force of the swirling
ideally, the models still capture enough of the fundamental physics of gas pushes the liquid to the wall of the pipe, where it may form an
the problem to allow interpolation and extrapolation to different upward-spiraling continuous ribbon of liquid.
fluid-flow conditions. CFD simulations predict details of the com- At present, the onset of liquid carry-over is predicted for low to
plex hydrodynamic-flow behavior in the GLCC, including flow field, moderately high gas rates. The key to onset of liquid carry-over has
holdup distribution, and trajectories of discrete particles of the dis- been to predict accurately the maximum liquid holdup (volume frac-
persed phase. While well-suited for local simulation of single-phase tion) occurring in the upper part of the GLCC under zero-net-liquid-
or dilute dispersion flows, current CFD simulators cannot yet handle flow conditions and its effect on the pressure balance between the gas
the complexities of the full range of multiphase flow. Furthermore, and liquid legs. Fig. 2 compares model predictions with experimen-
CFD models of large piping systems that include the GLCC typical- tal results in plots of the maximum liquid holdup in the upper GLCC
ly are too unwieldy to be practical for design purposes. region (i.e., zero-net-liquid-flow holdup, yL0, vs. the superficial gas
Because mechanistic models are greatly simplified, they are not velocity, vgs, in the GLCC).2 Additional data collected for a range of
as detailed, rigorous, or accurate as CFD models. However, mech- liquid viscosities from 1 to 10 cp showed negligible effect on the
anistic modeling has many advantages: speed of setup and compu- zero-net-liquid-flow holdup.6 Once the maximum liquid holdup
60 JULY 1998 •
100
vLs = 0.05 ft/sec
80 vLs = 0.1 ft/sec
d100, µm
60 vLs = 0.5 ft/sec
40
20
0
vtis/vz
decreases with increasing vtis /vz ratio up to about 100 and remains
approximately constant for larger values of this ratio.
Fig. 4—Bubble-trajectory comparison of mechanistic model and The region from the bottom of the vortex to the liquid exit is
CFX simulations with v Ls = 0.25 ft/sec, v gs = 10 ft/sec,
vtis/vz = 34, d= 3 in., and db = 20 µm.9 where small bubbles are separated and captured by the gas-core fil-
ament. Because vortex height is a strong function of tangential-inlet
velocity and bubble-trajectory length diminishes with vortex
allowed in the upper part of the GLCC is known for a given gas rate, height, an optimum tangential-inlet velocity must exist that mini-
the pressure-balance calculation is used to determine the liquid rate mizes gas carry-under. A tangential-inlet velocity that is too low
required to achieve this holdup and initiate liquid carry-over. produces insufficient centrifugal and buoyancy forces, whereas the
Fig. 3 compares the experimental and predicted operational available length for bubble trajectory is too short with a tangential-
envelopes for a 3-in. laboratory GLCC in a loop configuration, inlet velocity that is too high. As yet, a general scheme to determine
operated with air and water at low pressures.2 The operational optimum velocity has not been presented.
envelopes are presented in terms of superficial liquid velocity, vLs, Work is now in progress to develop the methodology to predict
vs. superficial gas velocity, vgs, in the GLCC. The agreement of overall separation efficiency in a GLCC. This requires two addition-
model predictions with the data is very good. Comparison with
data from Ref. 6 showed that the model seems to capture the phys-
ical phenomena and predict well the reduction of the operational
envelope with increasing liquid viscosity.
Future improvements to liquid-carry-over modeling will include
expansion to different operational conditions (e.g., high gas rates)
as well as prediction of the quantity of liquid carry-over and
dynamic responses to flow-rate surges.
Gas Carry-Under. Three mechanisms have been identified as possi-
ble contributors to gas carry-under in the liquid stream: (1) shallow
bubble trajectories prevent small bubbles from escaping to the gas-core
filament, (2) rotational-flow instability causes helical whipping and
breaking of the gas-core filament near the liquid exit, and (3) liquid-
rate surges can produce a concentrated cloud of bubbles that hinders
bubble migration to the gas core. Currently, attempts to predict gas
carry-under have focused only on the first mechanism, discussed next.
Bubble-Trajectory Analysis. This analysis is carried out by
assuming successive steady-state force-balance calculations on a
bubble. The forces acting on the bubbles are centrifugal, buoyancy,
and drag. Recent work compared bubble trajectories predicted by
the mechanistic model and CFD simulations for the same flow con-
ditions.9 Fig. 4, where x/d and r/R are the dimensionless axial and
radial coordinates below the GLCC inlet, respectively, provides an
example of such a comparison. The figure shows good agreement
with respect to the trend and absolute value.
Bubble-trajectory analysis10 was used to predict the onset of gas
carry-under and separation efficiency for different sized bubbles in
a manner similar to the liquid/liquid analysis carried out for hydro-
cyclones.11 The minimum diameter of the bubble that always
migrates from the GLCC wall to the gas core and thus is separated
(i.e., d100) was predicted. Fig. 5 shows the effect of the ratio of the
tangential velocity at the inlet slot to the axial velocity in the GLCC
(namely, vtis /vz) on d100. The continuous line represents the regres-
sion curve of the simulation results. For these conditions, d100
• JULY 1998 61
Rcap/R
6 in.
Vt, ft/sec
12 in.
vtis/vz
persed phase on the flow of the continuous phase and the interface
Fig. 6—Axisymmetric-tangential-velocity prediction vs. data for between the two phases. Recent two-phase-flow CFD simulation
a 7.5-in. GLCC operated with air and water at atmospheric con- work has proceeded on two fronts: with CFX14,17 and through
ditions.14 development of a dedicated internal code.17 The two-phase simula-
tions provided details of the velocity field and gas-void-fraction dis-
al pieces of information: the amount of gas entrained and the bub- tribution. The simulations also provided the free interface between
ble-size distribution. Coupling these to the bubble-capture efficien- the gas and liquid phases (vortex), which compared favorably with
cy ultimately will enable prediction of overall separation efficiency. experimental data. Fig. 8, which shows the gas-void-fraction distri-
bution in the GLCC, gives an example of the results obtained. The
CFD Simulation. Verifying mechanistic models with real data is figure reveals that the gas-void-fraction values at the top and bottom
not always practical or possible. CFD simulations are used to of the GLCC are nearly unity and nearly zero, respectively, indicat-
validate and improve the mechanistic models. CFD simulations ing efficient separation. For the first time, results have predicted the
for the GLCC can be lumped into two broad categories: single- gas-core-filament diameter accurately and provided insight into the
phase flow with particle tracking and two-phase flow. mechanism for its formation (continuous entrainment and radial
Single-Phase Flow and Particle Tracking. The simplest and migration of small gas bubbles into the gas core).*
most widely used approximation for CFD simulation of two-phase
flow is to consider single-phase flow populated with particles (bub- *Unpublished results, F.M. Erdal, U. of Tulsa, Tulsa, Oklahoma (1998).
bles) that neither interact with each other nor influence the flow.
This, in effect, is simply solving for a single-phase-flow field and
superimposing particle-trajectory tracking. α=0.98
CFD and bubble-trajectory analysis were used to investigate the
sensitivity of gas separation to bubble-size distribution.12,13 Two-
and three-dimensional (2D and 3D) simulations14 were carried out
with CFX, a commercially available CFD code.15 The authors con-
cluded that the axisymmetric simulations (2D) gave good results
compared with the 3D simulations. Fig. 6 compares single-phase
CFD simulations with experimental data.16 Both the data and CFD
simulations demonstrated that the tangential-velocity distribution
is dominated by a forced vortex, confirming this assumption in the
mechanistic models. Furthermore, the CFD simulations also veri-
fied the mechanistic model with respect to axial decay of tangen-
tial-velocity distribution (5 to 7% L/d decay).
The simulations in Ref. 14 also predicted the existence of an
axial-flow-reversal region where the flow is downward near the wall
and upward in the center core. The bubble-capture radius, Rcap, is
defined as the radial location where the axial-velocity component is
zero as the flow reverses from downward to upward. Bubbles that
migrate into the capture-radius area are separated and pushed
upward into the upper part of the GLCC. Fig. 7 shows the capture
radius as a function of the tangential-/axial-velocity ratio, vtis /vz, and
axial location below the inlet. The results indicate a rapid decline of
the capture radius as the velocity ratio decreases below 10. The cap-
ture radius and the reversal in the axial-velocity profile recently have
α=0.00
been incorporated into the mechanistic model.9
Two-Phase Flow. Actual two-phase-flow CFD simulation is still in Fig. 8—Void-fraction distribution for a 7.5-in. GLCC operated with
air and water at atmospheric conditions.14
its infancy. Such simulations should predict the influence of the dis-
62 JULY 1998 •
This type of measurement-loop configuration affords several
advantages over either conventional separation with single-phase
measurement or nonseparating multiphase meters. The loop con-
figuration is somewhat self-regulating, which can reduce or even
eliminate the need for active level control. The compactness of the
GLCC allows the measurement loop to weigh less, occupy less
space, and maintain less hydrocarbon inventory than a conven-
tional test separator. The advantages of a GLCC metering loop over
a nonseparating three-phase meter include much improved meter-
ing accuracy of individual phases over a wider range of flow rates
and significantly lower cost.
For flow conditions where gas carry-under cannot be prevented,
a three-phase metering system is required on the liquid leg. In gen-
eral, the accuracy of a multiphase meter on the liquid leg benefits
significantly from removal of some of the gas. Most multiphase
meters have an upper limit on the gas volume fraction allowed
through the meter to maintain their accuracy specifications. Apart
from improved accuracy, partial gas separation provides the addi-
Fig. 9—GLCC in a multiphase metering loop configuration. tional benefit of a smaller, less expensive multiphase meter. For
multiphase meters (whose price scales directly with size), the cost
APPLIC ATIONS savings of using a smaller meter in conjunction with a GLCC can
A variety of GLCC applications have requirements that may vary be four times the cost of the GLCC.
from partial to complete gas/liquid separation. Recent technologi- Partial Processing (Separation). A compact GLCC is often very
cal development has helped increase deployment of GLCC separa- appropriate for applications where only partial separation of gas
tor systems in the industry. from liquid is required. One such application is the partial separa-
tion of raw gas from high-pressure wells to use for gas lift of low-
Successful Applications. The GLCC modeling effort to date has pressure wells. The GLCC was a central feature in an offshore raw-
resulted in successful deployment of the GLCC in a variety of gas-lift system designed by Chevron that allowed elimination of gas
selected applications, as discussed next. compressor and lift-gas pipelines.19
Multiphase Measurement Loop. Most of GLCC’s deployed to Compact Separation Systems. Compact separation systems are a
date (approaching 100) have been configured in a multiphase key element in reducing cost of production operations through
metering loop. Fig. 9 is a schematic of the GLCC in a multiphase reduction of size and weight. Furthermore, separating a significant
metering loop, first introduced by Liu and Kouba,18 and Fig. 10
shows a GLCC field prototype operated by Chevron in Oklahoma.
• JULY 1998 63
ly equally to separation. Technological developments in both
GLCC hardware and software should reduce the size and improve
the performance of vertical separators. One challenge in optimizing
the size of a GLCC for production separation is designing a system
that can respond quickly to surges without serious upsets.
CONCLUSIONS
The GLCC is a compact, low-cost separator suitable for a wide range
of applications. The single biggest impediment to widespread imple-
mentation of the GLCC is the lack of proven performance-prediction
Fig. 11—Enhancement of existing conventional separator with a tools that are valid over a wide range of operating conditions. These
GLCC.1 tools are essential to reliable deployment of GLCC technology.
Performance-prediction tools based on empirical formulations are
limited in their ability to interpolate or extrapolate to new conditions.
portion of the gas reduces fluctuations in the liquid flow and may CFD simulations can capture much detail of local hydrodynamics but
result in improved performance of other downstream separation are too computationally intensive, time-consuming, and complicated
devices, such as a wellhead desanding hydrocyclone. Chevron is to apply to large systems. While CFD modeling is essential to improv-
investigating the series combination of a GLCC with a free-water- ing our understanding of the flow hydrodynamics in a GLCC, it is
knockout hydrocyclone and a deoiling hydrocyclone in an effort to impractical and therefore insufficient as a general design tool.
improve discharge-water quality. Mechanistic modeling is a reasonable compromise between the
The GLCC was used to control gas/liquid ratio of a two-phase- simplicity of empirical formulations and the complexity of CFD.
flow mixture entering a multiphase pump to improve pumping effi- Mechanistic modeling can be validated with CFD simulations to
ciency.20 Another study showed several combinations of GLCC and capture the fundamental physics of the flow without excessive
jet pumps that could be used to extract energy from high-pressure detail. The combination of CFD and mechanistic modeling pro-
multiphase wells to enhance production from low-pressure wells.21 vides a realistic approach to obtaining useful tools for design of and
Enhancement of Existing Separators. Cyclone separation performance predictions for the GLCC.
already has proved useful in internal separation devices for large New applications and improved designs are rapidly being
horizontal separators. The GLCC may also function as a useful explored and developed. Several companies are already using the
external preseparation device to enhance performance of existing GLCC for some applications (e.g., the multiphase metering loop)
horizontal separators (Fig. 11). By separating part of the gas, the and exploring new applications. The GLCC is now commercially
separator level might be raised to increase residence time without available through some vendors and is being evaluated by other
encountering the mist-flow regime in the vessel. Petrobrás Brazil companies. Deployment of GLCC’s in new and existing applica-
has retrofitted an existing separator in one of its fields with a GLCC tions will flourish as understanding of and confidence in predict-
preseparator.1 Another company is evaluating enhancement of ing flow behavior and performance of the GLCC matures.
their existing test separators with GLCC preseparation.
Commercial GLCC Products. Most GLCC’s to date have been field NOMENCL ATURE
fabricated for relatively straightforward applications. Applications of d= GLCC diameter, L, in.
and demand for GLCC’s are growing rapidly. Several vendor compa- db = bubble diameter, L, µm
nies are in the process of incorporating the GLCC into their com- d100= minimum bubble diameter that always migrates from
pact-separator product line. Also, as mentioned before, a commercial GLCC wall to gas core, L, µm
multiphase metering system that uses a GLCC and a second-stage L= length, L, ft
horizontal separator is now available. Greater commercialization will p= pressure, m/Lt2, psig
be needed to meet the growing industry demand. r= radial coordinate, L, ft
R= GLCC radius, L, in.
Future Applications. Current successful GLCC applications lend Rcap = bubble-capture radius, L, in.
confidence to future potential GLCC configurations. This requires vgs = superficial gas velocity, L/t, ft/sec
enhancement of the existing models and is currently under way. vLs = superficial liquid velocity, L/t, ft/sec
The following are two of the most compelling applications. vt = tangential velocity, L/t, ft/sec
Subsea Production. The biggest impact to the petroleum industry vtis = tangential velocity at inlet slot, L/t, ft/sec
from GLCC technology may be in subsea separation applications. vz = axial velocity, L/t, ft/sec
Conclusions in Ref. 22 state that “wellhead separation and pumping is yL0= zero-net-liquid-flow holdup, fraction
the most thermodynamically efficient method for wellstream transfer x= axial coordinate, L, ft
over long distances, particularly from deep water.” In a recent study, α= void fraction
Prado et al.23 argued that this is applicable to shallow and moderately
deep waters. Undoubtedly, development of marginal offshore fields will ACKNOWLEDGMENTS
depend on development of efficient and economical technologies. We thank Chevron Petroleum Technology Co. and the other mem-
Subsea applications require a high degree of confidence in separator bers of the Tulsa U. Separation Technology Projects (TUSTP) for
design and performance while demanding that the equipment be sim- supporting this work.
ple, compact, robust, and economical. Here again, the virtues of the
GLCC should place it in good standing among competing technologies. REFERENCES
Production Separation. Vertical separators with tangential inlets 1. Gomez, L.E.: “A state-of-the-art Simulator and Field Application
are fairly common in the oil field. These predecessors of the GLCC Design of Gas-Liquid Cylindrical Cyclone Separators,” MS thesis, U.
are often big and bulky, with perpendicular low-velocity tangential of Tulsa, Tulsa, Oklahoma (1998).
pipe inlets. The tangential velocities are usually so low that gravi- 2. Arpandi, I. et al.: “Hydrodynamics of Two-Phase Flow in Gas/Liquid
tational, centrifugal, and buoyancy forces contribute approximate- Cylindrical-Cyclone Separators,” SPE Journal (December 1996) 427.
64 JULY 1998 •
3. Kouba, G.E. and Shoham, O.: “A Review of Gas-Liquid Cylindrical SI METRIC CONVERSION FACTORS
Cyclone (GLCC) Technology,” paper presented at the 1996 Intl. IBC cp ↔ 1.0* E−01 =Pa.s
Conference on Production Separation Systems, Aberdeen, 23–24 April. ft ↔ 3.048* E−01 =m
4. Kouba, G.E., Shoham, O., and Shirazi, S.: “Design and Performance of in. ↔ 2.54* E+00 =cm
Gas-Liquid Cylindrical Cyclone Separators,” Proc., BHR Group psi ↔ 6.894 757 E+00 =kPa
Seventh Intl. Conference on Multiphase Flow, Cannes, France
(1995) 307. *Conversion factor is exact.
• JULY 1998 65