Study of Gas-Water Flow Inside of a Horizontal Passive Cyclonic Gas-Liquid Phase Separator System Using Displacement-Current Phase Tomography

Electrical capacitance volume tomography (ECVT) is a non-invasive tomography technique developed to measure multiphase flows (Warsito et al., 2007). Flexible capacitance plates distributed along the outside wall of the measurement region are used with an ECVT data acquisition system (DAS) to collect real-time measurements of inter-electrode capacitances. These capacitance measurements can then be used to derive information of the measured flow. An optimization reconstruction algorithm can also be employed to generate 3-D images of the sensing domain. ECVT has been used successfully in gas-liquid systems including a previous study on a PCGLS (Wang et al., 2016).

ECVT measurements are collected using alternating current (AC) signals. As such, the measured current has two components – the quadrature component (used to extract the capacitance) and the in-phase component. Typical ECVT measurements and images only utilize the quadrature component as the magnitude of the phase component is relatively small. However, this phase component contains valuable information as it depends on the conductivity (dielectric loss) present in the measurement region. Displacement-current phase tomography (DCPT) is developed to utilize this phase component to derive information and images of flows. DCPT is especially useful for lossy environments, such as a water continuous system, because of the relationship between phase and the systems dielectric loss. DCPT has been shown in previous studies to be better for imaging air-water systems than ECVT (Gunes et al., 2017). DCPT uses the same data collection system as ECVT and can be collected at the same time as capacitance measurements. This allows for both measurement methods to be utilized if needed (Gunes et al., 2017). In general, DCPT is best suited for systems containing at least one high-conductivity medium, like water, whereas ECVT is best suited for systems containing low-conductivity media, such as an oil-air flow.

The PCGLS is made out of a clear acrylic tube with an outer diameter (OD) of 3.5 in (8.89 cm) and an inner diameter (ID) of 3 in (7.62 cm). A 0.25 in (0.635 cm) ID injection nozzle is attached to the outer wall near the gas outlet. A swirl stopping baffle plate is placed 2 in (5.08 cm) from the liquid outlet end of the cylinder. The baffle plate is used to stop the vortex flow structure before it reaches the liquid outlet. The PCGLS cylinder length is about 9 in (22.86 cm), meaning the total vortex region length, between the gas outlet and baffle plate, is about 7 in (17.78 cm) long. The air outlet has an OD of 0.375 in (0.95 cm) and an ID of 0.125 in (0.32 cm). The water outlet has an ID of 0.875 in (2.22 cm). A basic diagram of the PCGLS can be seen in Figure 1.

Figure 1

An ECVT sensor with an inner diameter of 3.5 in (8.89 cm) and a length of 4.75 in (12.07 cm) is placed around the vortex region of the PCGLS. This sensor has 36 plates arranged in 3 layers, giving it 12 plates per layer. ECVT sensors have improved spatial resolution as the number of sensor plates are increased (Wang et al., 2015). The chosen sensor design (3 layers x 12 plates per layer) provides the best balance of axial and radial imaging resolution for this experiment. The data capture rate is 80 frames per second. Figure 2 is a photograph of the sensor placed around the described PCGLS.

Figure 2

The PCGLS is installed horizontally on a flow loop support frame. A water pump is used to circulate the water through the loop and PCGLS with a turbine flow meter placed inline after the pump to measure the water volumetric flow rate. A ball valve is used to control the water flow rate. The air is supplied by an air compressor with its volumetric flow rate measured and controlled using a rotameter.

The water and air are combined using a Y-connector and allowed a 22 in (55.88 cm) section of 0.75 in (1.905 cm) ID flow tubing to mix before being injected into the separator. After separation the water is recycled back into the system and the air is directed into the atmosphere. A ball valve is installed in the water outlet tubing to adjust the back pressure, thus controlling the gas core pressure and size inside the separator. The water outlet valve is adjusted to the point where all air is purged from the separator during water only injection and held constant through all experiments. Figure 3 shows a diagram of the PCGLS test flow loop.

Figure 3

The ECVT sensor is calibrated on the PCGLS before the experiments are conducted. For this process ECVT data is collected at an empty state and a full state. The empty state data corresponds to the sensing region being filled with the lowest permittivity material in the system, air in this case. The full state data corresponds to the sensing region being filled with the highest permittivity material, water in this case. This process calibrates settings within the DAS for optimal operation based on the flow materials. This also normalizes the data from the ECVT system between 0 and 1, corresponding to the empty and full states, respectively.

The PCGLS used in this experiment is relatively small compared to potential space applications of this system. The flow behavior of the liquid in the PCGLS is expected to vary as the section size is increased. At smaller section IDs, such as in this experiment, the wall friction has a higher impact on the velocity profile and phase distribution. However, the ECVT sensing technique is scale invariant as it operates in the quasi-static range of the spectrum. In this range, the excitation signal wavelength is orders of magnitude larger than the PCGLS size. For larger sizes, this measurement technique can still be used to study the impact of PCGLS size on the fluid dynamics in different size ranges.

All experiments are conducted at standard temperature and pressure. The water temperature is 20°C.

The air-water inlet liquid holdup experiments are conducted on a vertical flow loop at the NASA Glenn Research Center. The flow loop consisted of an up flow section and a down flow section with quick close valves (QCVs) in both sections to stop the flow for volume fraction measurements. A separate flow loop is used for this testing to achieve a better variety of liquid holdup and flow regimes and to better verify measurements using the QCVs. The loop is placed vertically to limit the effects of gravity on the flow and better simulate flow regimes seen in a zero gravity environment. Flow regimes observed in the vertical sections consisted of bubble, churn/transitional, annular, and slug flows which match the four main flow regimes observed in microgravity (Lowe and Rezkallah, 1999).

The flow loop test section is made out of polycarbonate tubing with a 1 in (2.54 cm) OD and 0.75 in (1.905 cm) ID. A 12 plate, 2 layer (6 plates per layer) ECVT sensor is placed around both the up flow and down flow test sections. For this test setup a 1 layer sensor could be used to adequately measure liquid holdup. However, 2 layers are used to provide additional data channels to make up for any faulty data points that may arise. The ECVT sensor is calibrated with air and water as the empty and full states, respectively, before experiments are conducted. The data capture rate is 125 frames per second. This sensor has a higher frame rate than the PCGLS because it has less plates. This means that the DAS has less plate pairs to activate and measure, increasing the speed at which the system collects data. Figure 4 shows a basic diagram of the test section used. Water flow rates ranging from 3.79 LPM to 18.95 LPM and air flow rates ranging from 1 SLPM to 50 SLPM are used during this experiment.

Figure 4

Under liquid only injection conditions air is purged from the system by the pressure placed on the liquid outlet. Thus, the liquid only injection flows consist solely of a rotating water layer with no air. Under all gas flow test conditions, a gas core forms in the center of the separator with a liquid layer forming along the edge of the separator. It is observed that, when an air-water mixture is injected, air bubbles appear in the water layer in the region between the mixture inlet and gas outlet. Beyond this region the bubbles combine with the gas core, creating distinct layers of gas and liquid. Figure 5 shows three photographs of the air and water layers under a constant water flow rate and three different air flow rates. It was observed that the air core may be off-centric, especially at lower air flow rates. Under the steady state conditions tested in this experiment, the liquid layer can be treated as a rigid body as it rotates around the gas core (Ellis, 2006). It is also observed that the gas core has a pulse like movement where it continuously expands and contracts. This movement is very small but becomes more noticeable as the air flow rate increases. It is possible that this pulse like movement is caused by the wave-like structure at the interface between the air core and water layer. This wave-like structure can be seen along the edges of the air cores in Figure 5. It is these irregularities that are utilized to calculate the rotational velocity of the liquid layer as outlined in the PCGLS Angular Velocity section below.

Figure 5

In order to determine the ECVT system accuracy, baseline separator features are measured without the ECVT sensor on the separator. The liquid holdup inside the separator is measured using a high speed video camera. The videos are analyzed to measure the radius of the gas core inside the separator. Distortion due to the water and acrylic tube are accounted for by comparing the size of the core to the size of the baffle plate using equation 1 where r_c is the air core radius, r_b is the baffle plate radius, and r_ci and r_bi are the measured radii of the core and baffle plate in the image.

Since the actual size of the baffle plate is known, the actual size of the core can be calculated easily by using the ratio of the size of the air core and baffle plate in the image. This measurement method assumes that the baffle plate is distorted equally to the air core. The measured gas core sizes are plotted against the air flow rate at different water flow rates, as can be seen in Figure 6.

Figure 6

Previous research has shown that the liquid holdup inside of the separator due to the gas injection rate follows a similar curve as the ones in Figure 6 (Hoyt, 2013).

To measure the rotational velocity of the water in the separator, a tachometer is used in conjunction with a cylindrical ABS (Acrylonitrile Butadiene Styrene) plastic rod, covered in reflective tape, placed inside the separator. Three sets of rotational velocity measurements with air injection are collected and averaged. The measurements are then plotted against the air flow rate, as can be seen in Figure 7.

Figure 7

As Figure 7 shows, the rotational velocity decreases initially when air is introduced (from 0 slpm gas flow rate to 2.3 slpm), then increases in a generally linear shape as the air flow rate is increased. Figure 8 shows the rotational velocity measurements plotted against water flow rate under liquid-only conditions. This graph suggests that the angular velocity is proportional to liquid flow rate under liquid-only injections, as is observed in previous research on similar separators (Wang et al., 2016; Ellis, 2006). A linear fit can be applied to derive the relationship between the water injection rate and the angular velocity inside the PCGLS. The linear relation for this study can then be expressed as (2)Ω (RPM)=59.26W(LPM)−183\Omega \,(RPM) = 59.26W(LPM) - 183 where Ω is the angular velocity of the water layer in rotations per minute and W is the water injection rate in liters per minute. As low water flow rates were not tested in this experiment it cannot be assumed that the linear fit would still apply under these conditions.

Figure 8

The use of a tachometer to measure the rotational velocity of the water layer serves two purposes in this study. The first purpose is to provide rotational velocity data from a secondary measurement technique to compare to the ECVT results. Second, the tachometer can be used to understand how the rotational velocity relates to the water flow rate under liquid-only injection conditions. The ECVT system utilizes irregularities in the interface between the gas and liquid layers to measure rotational velocity. In this experiment, there is no gas core under liquid-only injection conditions meaning that the sensor has no interface to measure the rotational velocity. The liquid-only injection curve developed from tachometer data is used in Equation 6 to quantify how the air and water injection flow rates affect the rotational velocity of the water layer.

The liquid holdup inside the test sections are measured visually using the QCVs. When the flow appears to reach a stable flow condition, data is collected with the ECVT system for 30 seconds. The QCVs are then closed and the water allowed to settle at the bottom of the test section. The liquid holdup is then calculated by measuring the height of the water and dividing by the total height of the test section. The QCV measurement process is conducted three times for each water and air inlet condition to get an average liquid holdup value.

The flow regimes observed in the up flow section consist of a bubble flow at low air flow rates that slowly become a pulsating churn flow at higher air flow rates. The down flow section also has a bubble flow at low air flow rates but this section quickly transforms into an annular/falling film flow as the air flow rate is increased. As such, a small change in the air flow rate is observed to cause a larger change in the down flow section than in the up flow section. This allows for lower liquid holdup values to be tested in the down flow section when the water flow rate is decreased and the air flow rate is increased.

Figure 9 shows a typical real-time DCPT reconstructed image for the liquid and gas distribution inside the separator. Several cross sectional slices perpendicular to the cylinder axis are shown. The reconstructed 2-D images are used to show the shape of the gas core and the liquid distribution within the PCGLS. In the figure, the gas core can clearly be seen as the black and red region. The water is represented by the light blue and white region, with the darker blue pixels representing a region with an air-water interface. These 2-D images can be used to analyze the change in gas core size at various inlet conditions. In Figure 9 the core in images A and B are off-center and smaller than the core in images C and D. This changing air core position and size reflects that the air core has size and position fluctuations along the length of the PCGLS.

Figure 9

Figure 10 shows reconstructed gas core images for a constant water flow rate and increasing air flow rates. The imaged air cores are slightly off center, which is more noticeable at lower air flow rates (images A and B). This becomes slightly less noticeable at higher air flow rates but the cores are still slightly off-center. This is reflected in Figure 5 where the cores are observed to be off-centric, more so at low air flow rates. These images can be analyzed to determine the gas core radius by measuring between the dark blue pixels. The measured gas core radii from the DCPT reconstructed images are summarized in Table 1. These reconstructed images can also be analyzed to determine the liquid holdup throughout the sensor. Table 2 compares the imaged liquid holdup to the expected liquid holdup based on Figure 6.

Figure 10

The liquid holdup can also be measured without the reconstructed image by directly using the data from the ECVT sensor. The real time liquid holdup is calculated by using a predetermined equation that relates the DCPT measurements to the liquid holdup. The air void fraction η_air can then easily be obtained from the liquid holdup η_lh as: (3)ηair=1−ηlh{\eta _{air}} = 1 - {\eta _{lh}}Table 3 shows the visually estimated liquid holdup compared with the DCPT determined liquid holdup at varying water and air flow rates.

The measurements in Table 3 have an average absolute error of 5.80% between the visually determined and DCPT determined liquid holdups (error calculated as the absolute value of the difference between visual and DCPT holdup measurements). Some sources of this error may include not fully accounting for image distortion in the visual measurement and the potential that either peak in the periodic shape was measured rather than the gas core size.

As can be seen in Figure 11, the liquid holdup inside the separator tends to follow a periodic shape. This shape is to be expected due to the pulsating of the gas core as described in the PCGLS section above. When the gas core pulses outwards, the liquid holdup decreases, and when the gas core pulses inwards, the liquid holdup increases, creating a periodic shape. This periodic fluctuation in the signal is much more noticeable at higher air flow rates, as was also discussed in the PCGLS section. It can also be seen that the liquid holdup signal is noisier at lower air flow rates. This is probably due to the smaller size of the gas core under these conditions. If the gas core is too small then liquid will exit from both the liquid outlet and the air outlet (Ellis, 2006). This action of liquid exiting through the air outlet makes the gas core less stable, causing the liquid holdup signal to appear noisier. As the gas core size increases, less water exits from the air outlet until the system becomes stable with only water exiting from the liquid outlet and only air exiting from the gas outlet. When the system reaches this stability point the gas core stabilizes, causing the liquid holdup signal to become less noisy.

Figure 11

The basic model for a PCGLS consists of a steady air core in the center of a rotating water layer. In reality, this air core is generally unsteady with pulsations and local flow structures in the liquid stream. DCPT can capture these irregularities in the air core by analyzing the signal variations from different capacitance pairs. The angular velocity of the local flow structure can thus be calculated through cross correlation of the signal variations. Plate pairs within the same layer with 60° separation are chosen for this angular velocity analysis. The number of frames between local maxima in a single plate pair signal are determined and used in conjunction with the data capture rate to determine the angular velocity of the water layer inside the separator.

Figure 12 shows the DCPT determined angular velocity values plotted against the tachometer determined values from Figure 7. It can be seen that both measurement methods, DCPT and Tachometer, give similar results for each air and water flow rate. It is also observed that at a constant water flow rate, the angular velocity tends to increase linearly with the air flow rate. A linear fit applied to the angular velocity data suggests that the relation between angular velocity and air flow rate, when air flow rate is greater than zero, is as follows: (5)Ω (rpm)=17.95*Q(SLPM)+B−40\Omega \,(rpm) = 17.95 * Q(SLPM) + B - 40 where Q is the air flow rate in SLPM and B is the angular velocity under liquid-only injection conditions. Equation 2 can be substituted for B in Equation 5, giving the relation between air flow rate, water flow rate, and angular velocity for this study as: (6)Ω (rpm)=−17.95*Q(SLPM)+59.26*W(LPM)−223\Omega \,(rpm) = - 17.95 * Q(SLPM) + 59.26 * W(LPM) - 223

Figure 12

A potential source of error in the DCPT measurements is noise in the data signal. The noise in the data signal can affect the system's ability to detect the air core irregularities that are used to calculate angular velocity. This noise will also affect instantaneous liquid holdup readings but can be mostly averaged out if data is collected for a long enough time.

Nineteen different combinations of air and water flow rates are tested. Seven of these conditions are tested in the up flow section and twelve are tested in the down flow section. The up flow combinations are tested three times and the down flow combinations are tested twice. The down flow combinations are only tested twice due to time constraints on access to the test flow loop. Once again the real time liquid holdup is calculated using a predetermined equation relating the ECVT measurements to the liquid holdup.

Table 4 compares the visual liquid holdup measurements, measured with the QCVs, with the ECVT liquid holdup measurements for the nineteen flow combinations. For simplicity the three up flow tests are averaged together and the two down flow tests are averaged together to get a single value for each flow combination in each flow direction. The absolute error and relative error between the visual and ECVT measurements are plotted together in Figure 13. The absolute error is calculated as the absolute value of the difference between the visual and ECVT measurements. The relative error is calculated using equation 7 below: (7)Relative Error (%)=|Visual−ECVT|Visual*100{\rm{Relative}}\,{\rm{Error}}\,(\%) = {{|Visual - ECVT|} \over {Visual}}*100

Figure 13

It can be seen that the ECVT measurements and visual measurements closely match with an average absolute error of 2.44% and an average relative error of 5.03%.

90% of the ECVT measurements are below 5% absolute error and 95% are below 10% relative error. There is some inherent error between the ECVT and QCV measurements as they are not necessarily measuring the same liquid holdup. The QCV measurements give an average liquid holdup by assuming that the liquid holdup is constant throughout the test region. In reality the liquid holdup varies across the test region and is constantly fluctuating as the water and air flow through. The ECVT system measures these fluctuations in real time meaning that the average ECVT measurement is an average of the liquid holdup fluctuations. This is why the absolute error is greater in tests where liquid holdup fluctuations were larger, specifically the high air flow tests within the up flow section.