LOCAL VELOCITY MEASUREMENTS AND COMPUTATIONAL FLUID DYNAMICS (CFD) SIMULATIONS OF SWIRLING FLOW IN A CYLINDRICAL CYCLONE SEPARATOR

Proceedings of ETCE 001: Engineering Technology Conference on Energy February 5-7, 001, Houston, Texas ETCE 001-17101 LOCAL VELOCITY MEASUREMENTS AND COMPUTATIONAL FLUID DYNAMICS (CFD) SIMULATIONS OF SWIRLING FLOW IN A CYLINDRICAL CYCLONE SEPARATOR Ferhat M. Erdal and Siamack A. Shirazi Department of Mechanical Engineering The University of Tulsa Tulsa, Oklahoma 74104 Tel. (918) 631-3001 Fax (918) 631-397 siamack-shirazi@utulsa.edu ABSTRACT Local measurements and 3-D CFD simulations in Gas- Liquid cylindrical Cyclone (GLCC ) separators are scarce. The main objective of this study is to conduct local measurements and 3-D CFD simulations to understand the swirling flow behavior in a cylindrical cyclone with one inclined tangential inlet. Axial and tangential velocities and turbulent intensities across the GLCC diameter (ID=3.5 ) were measured at 4 different axial locations (1.5 to 35.4 below the inlet) by using a Laser Doppler Velocimeter (LDV). The liquid flow rate was 7GPM, which corresponds to an average axial velocity of 0.73 m/s and Reynolds number of 66,900. Measurements are used to create color contour plots of axial and tangential velocity and turbulent kinetic energy. Color contour maps revealed details of the flow behavior. Additionally, 3-D CFD simulations with different turbulence models are conducted. Simulations results are compared to LDV measurements. INTRODUCTION The use of Gas-Liquid Cylindrical Cyclone (GLCC ) separators for gas-liquid separation is a new increasing technology for oil and gas industry. The GLCC separator is an attractive alternative to the conventional vessel-type separator, especially for offshore platforms in oil and gas production operations. As a result, it is essential to develop predictive tools for design and to be able to improve the technology of the GLCC. The GLCC configuration is shown schematically in Fig. 1. The gas and liquid mixture flows through an inclined inlet section, to enhance stratification, before reaching a tangential inlet slot. As a result of the tangential inlet, a vortex is formed causing the gas and liquid to separate due to the centrifugal/buoyancy forces. The liquid moves toward the wall and downward, while the gas flows to the center and exits from the top. For certain operating conditions, some liquid flows with the gas, and moves up toward the gas leg. This phenomenon is referred to as liquid carry-over. On the other hand, some gas may be entrained with the liquid and exit from the bottom of the GLCC, namely, gas carry-under. To develop mechanistic models to estimate the efficiency of the separator, information about details of flow such as velocity profiles and turbulent intensity is required. Up to date, there is no experimental data available on local measurements of axial and tangential velocities and turbulent kinetic energy in the GLCC. Several experimental studies on single-phase turbulent swirling flow in pipes, generated with tangential injection or guide vanes, have been reported. One of the first experimental studies in this area is by Nissan and Bressan [1]. To generate the swirling flow, water was injected through two horizontal tangential inlets. The flow field was measured with impact probes. The axial velocity distribution showed a region of flow reversal near the center of the tube. It was reported that under some circumstances there was visual evidence of double flow reversal, with water flowing forward near the wall and in the center of the cylinder and moving backwards in the region in between. Ito et al. [] investigated swirl decay in a tangentially injected swirling flow. They used water as the working fluid with a high ratio of tangential momentum to axial momentum, namely, 50. The measurements were made with a multielectrode probe. The tangential velocity distribution showed that there were two flow regions: a region of forced-vortex flow near the center of the tube and a surrounding region of freevortex flow. The swirl decayed with the axial distance, resulting in a decrease in the extent of the solid rotational flow (forced vortex). Turbulence in decaying swirling flow through a pipe was studied by Algifri et al. [3] using a hot-wire probe. Air 1 Copyright 001 by ASME

Multiphase Flow Gas Liquid Figure 1. Schematic of GLCC Configuration. was used as the working fluid and it was given a swirling motion by means of a radial cascade. The velocity profiles were presented three components of velocity. They found that for high swirl intensity the Reynolds number strongly affects the velocity distribution. It was suggested that the tangential velocity distribution, except in the vicinity of the pipe wall, can be approximated by a Rankine vortex, which is a combination of a free and a forced vortex. Kitoh [4] studied swirling flows generated with guide vanes. The flow field was measured with X-wire anemometers. It was shown that the swirl intensity decays exponentially. Yu and Kitoh [5] developed an analytical method to predict the decay of swirling motion in a straight pipe. They indicated that at lower Reynolds numbers the swirl appears to decay at a faster rate than for higher Reynolds numbers. Chang and Dhir [6] studied turbulent flow field in a tube where air is injected tangentially using a single rotated straight hot wire and single rotated slanted hot wire anemometers. Profiles for mean velocities in the axial and tangential directions, as well as the Reynolds stresses, were obtained. The axial velocity profile shows the existence of a flow reversal region in the central line of the tube and an increased axial velocity near the wall. Tangential velocity profiles have a local maximum, the location of which moves radially inwards with axial distance. The swirl intensity, defined as the circulation over a cross sectional area, was found to decay exponentially with axial distance. Computational fluid dynamics (CFD) codes allow the simulation of very complex flow problems without the expense of experimental setup and measurements. There are several previous CFD investigations of the flow behavior in the GLCC. Erdal et al. [7] presented CFD simulations utilizing a commercial code called CFX. The simulations included details of the hydrodynamic flow behavior in the GLCC, for both single-phase and two-phase flow. An expression was developed for an equivalent inlet tangential velocity for the axisymmetric model. A sensitivity study on the effects of the ratio of the inlet tangential velocity to the average axial velocity on the flow behavior in the GLCC was also carried out. Recently, Motta [8] presented a simplified CFD model for rotational two-phase flow in a GLCC separator. The model assumed an axisymmetric flow but considered three velocity components. The study also presented a comparison between the proposed model and predictions of a commercial CFD code (CFX). Erdal et al. [9] conducted bubble trajectory simulations in the GLCC by using a axisymmetric geometry. Their work revealed that turbulent dispersion has a significant effect on bubble trajectories and the percent bubbles that are carried under. The simulations with turbulent dispersion indicate that many bubbles migrate to the center of the GLCC to form the gas-core-filament below the gas-liquid free interface. The bubble trajectory results indicate that there is an optimal L/D ratio beyond which the percent bubbles that are carried under significantly decreases. Based on experimental data and theoretical studies performed at The University of Tulsa (Gomez et al. [10]), a design program has been developed which has mechanistic models to predict the operational envelop for liquid carry-over, bubble trajectories, and simple -D (axisymmetric) axial and tangential velocities. The models for axial and tangential velocities were developed by Mantilla et al. [11]. However, Mantilla et al. correlations were based on data for horizontal tangential inlets. Furthermore, Mantilla et al. [11] model was based on literature data for multiple (two and more) horizontal tangential inlets where the actual flow is axisymmetric. Several flow conditions and inlet geometries were used for the previous experiments in the literature. However, experimental data for one tangential inclined inlet geometry is scarce or does not exist. In addition, all the previous studies are based on axisymmetric flow condition and axismmetric CFD simulations. Therefore, the objective of this investigation is to gather local velocity data for flow inside the GLCC below the inlet for one tangential inclined inlet geometry and compare with 3-D CFD simulations. EXPERIMENTAL FACILITY AND MEASUREMENTS Since the flow is mostly liquid below the inlet of GLCC, an experimental facility for single-phase flow has been designed and constructed for obtaining measurements of axial and tangential velocities and turbulence quantities below the inlet of the GLCC. A schematic of the experimental facility is shown in Fig.. The experimental facility meets following requirements: Single-phase, closed liquid flow loop. Test section accommodates the Laser Doppler Velocimeter (LDV) measurement technique. Copyright 001 by ASME

Easy and quick change of different inlet configurations. Three phase, HP centrifugal pump, capable of producing 80 GPM (at max. head of 88 ft). Two rotameters (5-40 GPM) and flow rate measurement tree to measure flow rates for different inlet configurations (to be tested for future). A bypass line with a gate valve to control the flow rate through the test section. A gate valve on the outlet of the GLCC to increase pressure and to release air from the system. A 00 gallon reservoir to accommodate high flow rates. An air release bleed valve at the top of the GLCC. P Reservoir Bypass Line Inlet Section Rota Meters Cap for Two Inlet Configuration Air Bleed Valve Test Section LDV Probe Valve Outlet Section Figure. Schematic of Experimental Facility. Inclined Inlet Inlet Inclined Inlet Pipe 1.5 x LDV 1.5 Measurement Plane Flow Direction 3.5 4.8 Outlet 35.4 Top View Outlet Side View Figure 3. GLCC Test Section and LDV Measurement Plane. 3 Copyright 001 by ASME

ID PVC pipes are used for piping. Flow metering section is designed to allow flow rate measurements by rotameters for different inlet (one and two) configurations. Omega FL75-F variable area flow meter is used which has an accuracy of ±.5%. GLCC configuration is divided in three sections, namely, Inlet Section, Test (Measurement) Section, and Outlet Section as shown in Fig.. These three sections are connected to each other with flanges. One inclined inlet configuration is used in all experiments that are mentioned in this study. Inlet pipe was inclined 7 with respect to the horizontal plane. GLCC test section is made out of 3.5 (88.9 mm) ID clear acrylic pipe as shown in Fig 3. In order to maintain a reasonably constant water temperature (and thus viscosity), the water in the flow loop is maintained at 0-5 C. The measurement plane is one plane between 1.5 to 35.4 below the inlet, as shown in Fig. 3. 4 measurement locations are selected in the measurement plane. Locations of these measurement lines are listed in Table-1. At each measurement location, axial velocity, tangential velocity and turbulent quantities are measured along the diameter by the LDV. Local measurements are conducted for flow rate of 7 GPM, which average axial velocity corresponds to 0.731 m/s and Reynolds Number is 66,900. The data is obtained using a two-component LDV developed by TSI, Inc., that includes the following components: a 3-Watt Argon-ion laser, a multicolor beam separator, a fiber optic probe for transmitting and receiving, a multicolor receiver, a digital processor, data storage and analysis software, and a two-axis traverse. For this study, a lens with a focal length of 36.6 mm and a focal distance of 350 mm is used with a probe beam spacing of 50.0 mm. Thus, the half-angle of the beams, κ, equals 3.95. Table 1 - Measurement Locations # L (in.) L (mm) L/D # L (in.) L (mm) L/D 1 1.5 317.5 3.6 13 18.4 466.7 5.3 13.0 330. 3.7 14 18.8 477.8 5.4 3 13.5 34.9 3.9 15 19.6 498.5 5.6 4 13.9 35.4 4.0 16 0.6 5.3 5.9 5 14.4 365.1 4.1 17 1.5 546.1 6.1 6 14.9 377.8 4.3 18.4 569.9 6.4 7 15.4 390.5 4.4 19 3.4 594.4 6.7 8 15.9 403. 4.5 0 4. 614.4 6.9 9 16.4 415.9 4.7 1 7.0 685.8 7.7 10 16.9 48.6 4.8 9.8 757. 8.5 11 17.4 441.3 5.0 3 3.6 88.7 9.3 1 17.7 449.3 5.1 4 35.4 900.1 10.1 To reduce the effects of the pipe curvature, a water-filled, clear-acrylic surrounding box is used. This enables the location of the measurement volume for each component (axial and tangential) to be approximately equal. The flow is seeded using silicon carbide particles with a mean diameter of µm. Approximately 1 cm 3 of particles are used for a 00-gallon reservoir of water. 50,000 data points are obtained at each point of all locations across the pipe, and the data points that lie outside of a three-standard deviation range from average are eliminated. Turbulent kinetic energy (k) is a measure of the kinetic energy per unit mass associated with the velocity fluctuations in three orthogonal directions is defined by Equation (1). 1 k = ( u ) + ( v ) + ( w ) (1) The LDV system employed in the study is a two-component system. Therefore, axial and tangential velocity fluctuations can be measured. Measurements showed that fluctuations in axial and tangential directions have the same order of magnitude. In order to get an estimate of k, radial velocity fluctuations are assumed to be average of the axial and tangential velocity fluctuations. Radial velocity fluctuations are calculated by using equation (). 1 ( v ) (u ) + (w ) () EXPERIMENTAL RESULTS Experiments are conducted for one inclined inlet configuration. Axial and tangential velocities and turbulence quantities are measured with LDV for flow rate of 7 GPM. Fig. 4 shows the normalized axial velocity profile at 1.5 below the inlet. At this location, there is a high downward axial velocity about 3.5 times higher than the relative average axial velocity. This high axial velocity is decreasing toward the center of the GLCC. On the other side, data shows a small region of negative axial velocities. This region represents flow reversal region where the flow is opposite to main flow direction. Near the other wall, data shows a small region of positive axial velocity. From this figure, it is clear that flow is not axisymmetric for the case of one tangential inclined inlet. Tangential velocity profile at 1.5 below the inlet is plotted in Fig. 5. Due to the rotation of the flow, tangential velocity is negative on one side and positive on the other side. This plot shows high tangential velocity near the wall regions. This high tangential velocity decreases toward the center, but location of the zero tangential velocity is off the center of the cylinder. On the left, tangential velocity starts with a flat profile and it takes a forced vortex profile near the center. Forced vortex profile is observed close to the region where tangential velocity changes sign. Fig. 6 shows measured turbulent kinetic energy (Equations 1 and ) profile at 1.5 below the inlet of the GLCC. Turbulent kinetic energy (k) at this location is relatively high with respect to U av. Data shows high k on 4 Copyright 001 by ASME

-1-0.5-45 -35-5 -15-5 5 15 5 35 45 0 U/U av D (mm) 0.5 1 1.5.5 3 3.5 4 7gpm Figure 4. Axial Velocity Profile at 1.5 Below The Inlet. -45-35 -5-15 -5-0.5 5 15 5 35 45 W/U av 4 3.5 3.5 1.5 1 0.5 0-1 -1.5 - -.5-3 -3.5-4 -4.5 D (mm) Figure 5. Tangential Velocity Profile at 1.5 Below The Inlet. k/u av 1. 1 0.8 0.6 0.4 0. 7gpm 7gpm 0-45 -35-5 -15-5 5 15 5 35 45 D (mm) Figure 6. Turbulent Kinetic Energy Profile at 1.5 Below The Inlet. the left side, near the wall region. Since this location is close to the inlet and the left side is on the inlet side of the GLCC, this high k can be contributed to the inlet effects. High k decreases towards the center. After the center it starts increasing again and it peaks about the location where tangential velocity almost zero and axial flow reversal occurs. It starts decreasing again after this location towards the wall. Measurements are repeated for rest of the 3 measurement locations along the diameter of the GLCC. Measured axial velocity profiles are shown in Fig. 7 for flow rate of 7 GPM. These profiles show that high positive axial velocity (downward flow) region and negative axial velocity (upward flow) region starts to disappear with respect to the measurement location. At 14.4 below inlet, the upward flow region completely vanishes. After this location downward flow (positive axial velocity) region can be observed near the wall on the right side of the pipe while upward flow starts to emerge on the left. At 16.7, upward flow can be observed in the axial velocity profile on the left. After this location, upward flow region and downward flow region develop more for a distance and start to vanish again at 18.6. At 0.6, up ward flow completely disappear again. Repetition of the similar behavior of upward and downward axial velocity can be seen in the measured axial velocity profiles in further down locations. As one can observe, it is difficult to conclude how the flow behaves from these individual plots. Therefore, contour plots of these measurements were created. Fig. 8 shows a color contour plot of axial and tangential velocities and turbulent kinetic energy that is generated from the local measurements. Measured axial and axial velocities and turbulent kinetic energy profiles at 4 locations in the measurement plane are used to obtain the color contour plots. The axial velocity contour plot clearly shows regions of high and low axial velocities and upward flow (negative axial velocity) regions. The blue and purple regions that have a helical shape represent the regions of low and negative axial velocities. In this helical shape region, appearance and disappearance of upward flow in the measurement plane can be observed. Downward spiral flow near the wall has high axial velocities. Strengths of both upward and downward flow decay as the flow moves toward the outlet. This decay appears to cause a stretch on the vortex (the wave length of the helical vortex increases with axial distance). Color contour plot of tangential velocities in Fig. 8 shows that tangential velocity is positive on one side (left) and negative on the other side (right). This is due to the rotation of the flow. Positive velocities represent the tangential velocity out of the page and negatives represent flow into the page. The tangential velocity is high near the wall region and it decays towards the center. There is a decay of tangential velocity in the axial direction towards the outlet or downward. Location of zero or low tangential velocity has a helical path similar to the one observed in the axial velocity contours. 5 Copyright 001 by ASME

) D ( m m ) s) D ( m m ) Calculated turbulent kinetic energy (k, Equations 1 and ) profiles, normalized with U av, are also color contour plotted in Fig. 8. This figure shows a relatively high k near the wall at the top. This high k decays downward in the near wall region. This is due to inlet effects and the inlet jet that is spiraling down into the cylinder. The high k region at the center which has a helical shape doesn t show a strong decay. Although axial and tangential velocities in this region are low with respect to near wall region, k is considerably high, comparable to average axial velocity. This high turbulence shows the instability of the flow at the center. This might have great impact on the separation of small bubbles below the inlet of GLCC. ( ) m/s) 1.5 13.0 13.5 13.9 14.4 14.9 15.4 15.9 16.4 16.9 17.4 Flow Direction 18.4 18.8 19.6 0.6 1.5.4 3.4 4. 7.0 9.9 3.6 17.7 (7 GPM) 35.4 Figure 7. Measured Axial Velocity Profiles (7 GPM). x 1.5 Flow Direction 350 400 450 500 550 U/U av 3.5.5 1.5 35.4-0.5 X (mm) 600 650 750 800 350 400 450 500 550 X (mm) 600 650 700 0.5 700 750 800 W/U av 4.0.7 1.3 0.0-1.3 -.7-4.0 350 400 450 500 550 X (mm) 600 650 700 750 800 k/u av 1.5 1. 0.9 0.7 0.4 0.1 850 850 850 900 900 900 (Not to Scale) -41 0 41 R (mm) -41 0 41 R (mm) -41 0 41 R (mm) Figure 8. Color Contours of Axial and Tangential Velocities and Turbulent Kinetic Energy (7 GPM). 6 Copyright 001 by ASME

x 1.5 35.4 Flow Direction (Not to Scale) 1.5 13.0 13.5 13.9 14.4 14.9 15.4 15.8 16.4 16.7 17.4 17.7 18.4 17.4 17.7 18.4 16.7 16.4 Pipe Wall 15.8 15.4 14.9 14.4 13.9 13.5 Figure 9. Construction of Horizontal Plane Color Contour Plot of Axial and Absolute Tangential Velocities (7 GPM). U/U av 3.5.5 1.5 0.5-0.5 1.5-18.4 Below the Inlet 13 1.5 W/U av 4 3 1 0 Figure 10. Horizontal Plane Color Contour Plot of Axial and Absolute Tangential Velocities at 1.5-18.4 Below the GLCC Inlet (7 GPM). In order to shed more light to the behavior of the flow and the helical shape that is observed, flow field in a horizontal plane is approximately constructed by using the measured velocity planes and knowing that the flow has a periodic shape. By neglecting the viscous effects, LVD measurements of axial velocity profiles, from 1.5 to 18.4 below the inlet, are processed to obtain approximate axial plane color contour plot of axial velocity for 7 GPM, as shown schematically in Fig.9. By considering the rotation of the flow, measured axial velocity profiles are placed in horizontal plane. Since the measurements are taken about every 0.5, 13 axial velocity profiles are placed 15 apart in a horizontal plane. Fig. 10 represents the color contour plot of the axial and absolute tangential velocities in a horizontal plane approximately at 1.5-18.4 below the GLCC inlet for 7 GPM. Black line along the diameter represents the measurement plane at 1.5 below the inlet, the location where the axial velocity profiles are obtained. In this plane, positive axial velocities represent downward flow, into the page. Negative axial velocities represent upward, flow out of the page. This plot clearly shows locations of different flow regions with respect to the measurement plane. While high axial velocities are on the left, low axial velocities and even some flow reversal (upward flow) exist on the other side, off the center of the cylinder. This also shows that the location of the different flow regions with respect to measurement plane can have great effect on the shape of measured axial velocity profiles (see Figs. 4 and 7). The color contour plot of the absolute values of tangential velocity in a axial plane approximately at 1.5-18.4 below the GLCC inlet for 7 GPM is also shown in Fig. 10. In this plane, positive velocities represent rotational or tangential velocity tangent to the location in counter clockwise. Clearly, the location of low tangential velocities is off the center of the cylinder. There is a high tangential velocity region near the right wall. The color contour plots of axial velocities in various axial regions are presented in Fig. 11 for 7 GPM. These plots show the rotation of the flow as well as the location of the different flow regions (downward and upward) associated with its horizontal plane. Because of the rotation, velocity profile shape changes with respect to measurement plane. Therefore, different axial velocity profiles are observed in measurements as seen in Fig. 7. This also suggests that the appearance and disappearance of upward flow in the axial velocity profiles are caused by the rotation of the flow. This rotational behavior explains the helical path observed in the measurements. Flow completes a 180 rotation from 1.5 to 18.4 for 7 GPM (Re = 66900. Color contour plot of axial velocity at horizontal plane is also used to calculate the flow rate, which can be used to estimate the accuracy of the horizontal plane approximations and error in measurements. The flow rate is obtained by performing an area integral over the horizontal plane. The calculated flow rate is 70.84 GPM which is close to the measured value by the rotameter (corresponds to 1.6% difference). 3-D CFD SIMMULATIONS Computational Fluid Dynamics (CFD) simulations were carried out to evaluate the effectiveness of using a CFD code to predict the complex swirling flow in the GLCC. For the present CFD simulations, a commercially available CFD code called CFX by AEA Technology, 1997 [1], is used. A 3-D GLCC geometry similar to the one used in the experimental work with one inclined tangential inlet is created using the CFD code grid generation program. The geometry is constructed with 300,000 computational grids. Inlet flow velocity is set to accommodate 7 GPM of water flow rate. Flow is first simulated with k-ε turbulence model with Hybrid dicretisation scheme. After numerical solution convergence was obtained, simulation was repeated with available higher order dicretisation schemes, Higher Upwind and QUICK respectively, by starting from previous simulation as initial conditions. The simulation results using higher order discretisation schemes were very similar to the results obtained using the Hybrid scheme. Simulations are repeated with Reynolds Stress turbulence model (RSM) with 7 Copyright 001 by ASME

the Hybrid dicretisation scheme. When the simulations were repeated using higher order differencing schemes (Upwind and QUICK), a converged numerical solution could not be attained. Measured axial velocity color contour plot at the measurement plane is compared to CFD simulations with the k- ε and RSM as shown in Fig. 1. Simulation results capture general rotational behavior and upward flow region. However, simulations with both turbulence models failed to predict the finer details of the flow such as upward flow region, the wavelength of vortex, and local axial velocity profiles. Simulation with the k-ε model predicts stronger upward flow regions and shorted vortex wavelength. Simulation with RSM predicts weaker upward flow with much longer wavelength of the vortex. In addition, RSM model predicts much stronger decay in the upward flow region. In Fig 13, CFD prediction of Tangential velocity color contour plots are compared to LDV measurements. Simulations with the k-ε model predicts higher tangential velocity than the measurements which indicates higher rotational flow. Simulation with the RSM predicts much lower tangential velocity than measurements. A comparison of Figures 1 and 13 show that there is a close correlation between predicted axial and tangential velocities. Fig. 14 shows comparison of axial velocities in a horizantal plane 13.5 below the inlet for different turbulence models. Both simulations cannot predict the location of the upward flow region. The k-ε model predicts more off the axis. RSM model predicts more close to the center. 1.5 13.5 14.4 U/U av 3.5.5 1.5 0.5-0.5 16.4 17.4 18.4 Figure 11. Horizontal Plane Color Contour Plots of Axial Velocity (7 GPM). 8 Copyright 001 by ASME

x 1.5 350 450 400 500 550 LDV Data U (m/s) 3.0.4 CFX (k-ε) U (m/s).46 CFX (RSM) U (m/s).46 Flow Direction 35.4 X (mm) 600 650 700 750 800 1.8 1.3 0.7 0.1 1.86 1.5 0.64 0.04 1.86 1.5 0.64 0.04 850-0.5 <-0.5 <-0.5 900 (Not to Scale) -41 0 41 R (mm) Figure 1. Measured Axial Color Contour Plot Comparison with CFD Simulations (7 GPM). x 1.5 350 450 400 500 550 LDV Data W (m/s) 3.0.0 CFX (k-ε) W (m/s).08 CFX (RSM) W (m/s).08 35.4 X (mm) 600 650 700 750 800 1.0 0.0-1.0 -.0 1.04 0.00-1.04 -.08 1.04 0.00-1.04 -.08 850-3.0 <-3.00 <-3.00 900 (Not to Scale) -41 0 41 R (mm) Figure 13. Measured Tangential Color Contour Plot Comparison with CFD Simulations (7 GPM). 9 Copyright 001 by ASME

U (m/s) 3.0.3 1.6 0.9 U (m/s).46 1.86 1.5 0.64 U (m/s).46 1.86 1.5 0.64 13.5 Below the Inlet LDV Data 0. -0.5 13.5 Below the Inlet CFX (k-ε) 0.04 <-0.5 13.5 Below the Inlet CFX (RSM) 0.04 <-0.5 Figure 14. Measured Axial Color Contour Plot Comparison with CFD Simulations (7 GPM) in a Horizontal Plane at 13.5 Below The Inlet. SUMMARY AND CONCLUSION Axial and tangential velocities and turbulent intensities across the GLCC diameter were measured at 4 different axial locations (1.5 to 35.4 below the inlet) by using a Laser Doppler Velocimeter (LDV). The liquid flow rate was 7 gpm, which corresponds to an average axial velocity of 0.73 m/s. Axial velocity measurements indicate the flow is highly nonsymmetrical around the axis of the cylinder. Axial velocity measurements also indicated downward flow near the wall of the cylinder and small upward flow near the center, but off the axis of the test section. The axial velocity contour map obtained from local measurements shows that the reverse axial velocity region (upward flow region) has a helical shape. The axial downward flow has a helical path near the wall. Tangential velocity measurements show that free vortex occurs near the wall region and forced vortex occurs near the center of the cylinder. The location of zero tangential velocity is also off the axis of the cylinder. Turbulent kinetic energy is relatively high near the center and in the region near the inlet. Turbulent intensities in different directions have similar profiles and same order-of-magnitude. The experimental results are compared with 3-D CFD simulations. In this simulation, k-ε and RSM turbulence models are used. The CFD simulations show the general trend of the experimental data, but fails to predict correct wavelength of the vortex and local axial and tangential velocity profiles. Simulation with the k-ε model predicts higher rotational flow. On the other hand, simulation with RSM predicts a high decay in the intensity of tangential and axial velocities. ACKNOWLEDGMENTS This work is supported by the member companies of Tulsa University Separation Technology Projects (TUSTP).. Also, the support of the Turkish Ministry of Education for the Graduate Research Assistant, Ferhat Metin Erdal, is much appreciated. REFERENCES [1] Nissan, A.H. and Bresan V.P., 1961, Swirling Flow in Cylinders, A. I. Ch. E. Journal, 7, No.4, pp.543-547. [] Ito, S., Ogawa, K., and Kuroda, C., 1979, Decay Process of Swirling Flow in a Circular Pipe, International Chemical Engineering, 19, No. 4, pp. 600-611. [3] Algifri, A.H., Bhardwaj, R.K., and Rao, Y.V.N., 1988, Turbulence Measurements in Decaying Swirl Flow in a Pipe, Applied Scientific Research, January, 45, pp. 33-50. [4] Kitoh, O., 1991, Experimental Study of Turbulent Swirling Flow in a Straight Pipe, J. of Fluid Mechanics, 5, pp. 445-479. [5] Yu, S.C.M. and Kitoh O., 1994, General Formulation for the Decay of Swirling Motion Along a Straight Pipe, International Communications in Heat and Mass Transfer, 1, No. 5, pp. 719-78. [6] Chang, F. and Dhir, V. K., 1994, Turbulent Flow Field in Tangentially Injected Swirl Flows in Tubes, Int. J. Heat and Fluid Flow, 15, pp. 346-356. [7] Erdal, F.M., Shirazi, S.A., Shoham, O., and Kouba, G., 1997, CFD Simulation of Single-Phase and Two-Phase in Gas- Liquid Cylindrical Cyclone Separators, SPEJ,, pp. 436-446. [8] Motta, B.R., Erdal, F.M., Shirazi, S.A., Shoham, O., and Rhyne, L.D., 1997, Simulation of Single-phase and Two-Phase Flow in Gas-Liquid Cylindrical Cyclone Separators, proceedings of the ASME Summer Meeting, Fluid Eng. Division, Vancouver, Canada, June -6. [9] Erdal, F.M., Mantilla, I., Shirazi, S.A, and Shoham, O., 1998, CFD Study of Bubble Cary-Under in Gas-Liquid Cylindrical Cyclone Separators paper SPE 49309, presented at 10 Copyright 001 by ASME

the 1998 SPE Annual Technical Conference and Exhibition New Orleans, LA, Sep. 7-30. [10] Gomez, L., Mohan, R., Shoham, O., and Kouba, G., 1998, Enhanced Mechanistic Model and Field Application Design of Gas-Liquid Cylindrical Cyclone Separator, SPE 49174, presented at the SPE 73 rd Annual Meeting, New Orleans, September 7-30. [11] Mantilla, I., Shirazi, S., and Shoham, O., 1998, Flow Field Prediction and Bubble Trajectory Model in GLCC Separators, ASME J. Energy Resources Technology, 11, March, pp. 9-14. [1] CFX-4., 1997, CFX 4. Flow Solver, AEA Technology, Oxfordshire, UK. 11 Copyright 001 by ASME