Liquid film characterization of horizontal, annular, two-phase, gas-liquid flow using time-resolved LASER-induced fluorescence

Transcription

1 Liquid film characterization of horizontal, annular, two-phase, gas-liquid flow using time-resolved LASER-induced fluorescence P.S.C. Farias 1, F.J.W.A. Martins 1, L.E.B. Sampaio 2, R. Serfaty 3, L.F.A Azevedo 1 1: Department of Mechanical Engineering, PUC-Rio, Rio de Janeiro, Brasil, Lfaa@puc-rio.br 2: Department of Mechanical Engineering - LMTA/PGMEC, UFF, Rio de janeiro, Brasil 3: Petrobras R&D Center, Rio de janeiro, Brasil Abstract A non-intrusive optical technique was developed to provide time-resolved longitudinal and cross sectional images of the liquid film in horizontal annular flow of air and water, revealing the interfacial wave behavior. Quantitative information on the liquid film dynamics was extracted from the time-resolved images. The planar laser induced fluorescence technique was utilized to allow for the optical separation of the light emitted by the film from that scattered by the air-water interface. The visualization test section was fabricated from a tube presenting nearly the same refractive index as water, what allowed the visualization of the liquid film at regions very close to the pipe wall. Longitudinal images of the liquid film were captured using a high speed digital video camera synchronized with a high repetition rate laser. Data sets were collected with sampling camera frequencies ranging from 250 to 3000 Hz. An image processing algorithm was employed to automatically detect the position of the air-water interface in each image frame. The thickness of the liquid film was measured at two axial stations in each processed image frame, providing time history records of the film thickness at two different positions. Wave frequency information was obtained by analyzing the time-dependent signals of film thickness for each of the two axial positions recorded. Wave velocities were measured by cross-correlating the amplitude signals from the two axial positions. The results obtained allowed the verification of the variation of the liquid film characteristics with global flow parameters, such as the liquid and gas flow rates. For the film cross section observations, two high speed digital video cameras were used in a stereoscopic arrangement. The laser sheet was mounted so as to illuminate a pipe cross section. Images from the left and right cameras were distorted by the use of a calibration target and an image correction algorithm. Distorted images from each camera were then joined to yield the complete instantaneous cross section image of the liquid film. Comparisons with results from different techniques available in literature indicate that the present technique presents equivalent accuracy in measuring the liquid film properties. Time resolved images of longitudinal and cross section views of the film were recorded, what constitute valuable information provided by the technique implemented. 1. Introduction In horizontal annular flow the liquid flows as a thin, non-uniform film around the tube walls, while gas flows in the pipe core. The liquid film in this flow regime presents a wavy structure formed by ripples on the base film moving at low velocities and larger and faster disturbance waves. Droplets entrained in the gas flow are also part of liquid transport. The measurement and prediction of the time-varying, non-uniform liquid film distribution around the pipe perimeter that characterizes this class of flow has been one of the main focuses of research in the literature, together with the prediction of pressure losses along the pipe. One of the key fundamental questions related to the liquid film behavior is the mechanisms that maintain the film at the upper pipe wall, compensating drainage induced by gravity. Several mechanisms have been proposed as indicated, for example, in the works of Butterworth & Pulling (1972) and Jayanti et al. (1990). These mechanisms are (i) secondary gas motion induced by the circumferentially varying film thickness, (ii) liquid entrainment and re-deposition, (iii) wave spreading due to the distortion of liquid film waves at the bottom of the tube and (iv) pumping action associated with the flow of gas over the disturbance waves. Measurements of the wave structure of the liquid film have been conducted for both, vertical and horizontal annular flows for many years. These measurements include local time variation of film thickness, wave velocity and frequency, as well as spectral properties of film thickness time records. In these studies, resistance probes (Jayanti et al., 1990 and Paras & Karabelas, 1991) and, - 1 -

2 more recently, optical methods were employed (Shedd & Neweel, 1998). Flow visualization has also been used as a tool to aid in characterizing the film wave behavior. Its different implementations have followed the development of image technology along the years. Ciné movie with steady external illumination was used in conjunction with dye injection by Taylor & Nedderman (1968) and by Butterworth & Pulling (1972). Later, high speed video systems were introduced and replaced the time-consuming movie processing (Hewitt et al., 1990). Sutharshan et al. (1995) employed the photochromic dye activation technique to generate fluid tracers within the liquid film that were followed by high-speed digital imaging equipment with external back lighting. The analysis of the digital images provided qualitative information on the effects of wave passage on the liquid axial and circumferential velocity in the film. Typical liquid films in horizontal annular flow range from a few micrometers to a few millimetres in thickness. Visualization of such small dimensions at the neighborhood of a solid wall is a challenge for optical techniques. Hewitt et al. (1990) used a tube material that presented nearly the same index of refraction as water, what allowed for the clear visualization of the film structure close to the wall. Rodríguez & Shedd (2004) employed the same tube material in a visualization setup based on the planar laser induced technique PLIF. In this technique a fluorescent dye dissolved in water and excited by a pulsed planar sheet of laser light is used to produce clear images of the air-water interface. The reflections from the interface are blocked by an optical filter placed in front of the digital camera, producing clear instantaneous images of longitudinal sections of the film. Attempts to measure liquid velocity within the film were made by Vassalo (1999) using hot film probes in vertical annular flow. More recently, Koplin (2004) obtained partial success with particle image velocimetry (PIV) and particle tracking techniques to estimate velocity field in the film. The present work is part of an ongoing research project aimed at providing simultaneous, timeresolved qualitative and quantitative information on the liquid film structure in horizontal annular flows. The experimental technique implemented builds on the previous works of Hewitt et al. (1990) and Rodríguez & Shedd (2004) in the sense that employs index of refraction matching techniques associated with PLIF. High frame rate cameras synchronized with high-repetition rate lasers were used to provide time-resolved, quality images of longitudinal and cross section views of the liquid film around the pipe perimeter. In the case of cross section views, two cameras were used in a stereoscopic arrangement. Calibration targets were used to distort the images obtained from the side viewing of the pipe cross section. Instantaneous images of the film were processed to enhance contrast and to extract the time dependent properties of the liquid film such as, film thickness, timeaveraged and RMS values of film thickness, frequency power spectra, wave velocity, and histograms of film thickness distributions. In the next sections a description of the technique developed is presented, together with results of experiments conducted in a test section specially constructed for this purpose. 2. Experimental Facility The present study employed non-intrusive optical techniques to provide time-resolved images of the liquid film in horizontal annular air-water flows. Two different setups were utilized, one for capturing side views of the lower portion of the liquid film and a second setup for obtaining instantaneous images of the complete cross section at a particular axial position of the liquid film. In both cases, the planar laser induced fluorescence technique PLIF was employed to allow for the optical separation of the light emitted by the film from that (more intense) scattered by the airwater interface (Rodríguez & Shedd, 2004). Rhodamine B at a concentration of 500 µg per liter of water was employed as the fluorescent material. The fluorescent material was excited by a sheet of green light (527-nm wavelength) emitted by a double cavity, high-repetition rate, Nd-YLF laser. The liquid film in annular horizontal flows is expected to display small thicknesses, ranging from a - 2 -

3 few micrometers to thousands micrometers. Optical distortions due to the mismatch between the indexes of refraction of the tube wall material and the liquid, preclude the correct visualization of the film thickness. For this reason, the test section designed for the experiments employed pipes fabricated from FEP (Fluorinated Ethylene Propylene), which has nearly the same refractive index as water, what allowed for the visualization of the liquid film at regions close to the pipe wall (Hewitt et al., 1990). Figure 1 presents a schematic view of the test section utilized in the experiments for measuring the liquid film at the lower part of the tube. Water from a pump was fed to a 15-mmdiameter and 255-diameter-long FEP tube. Air was supplied to the test section by a centrifugal compressor. Air and water were mixed at a tee connection located at the inlet section of the tube. Calibrated rotameters were used to measure the air and water flow rates. The air-water flow exiting the tube was directed to a centrifugal separator from where the water was returned to the pump inlet, while the air was vented out of the laboratory space. A rectangular visualization box was installed at a distance of 190 diameters from the inlet. The box was filled with water in order to minimize optical distortions due to the pipe wall curvature. A Pegasus dual-cavity, Nd-YLF, high-repetition laser provided illumination of a longitudinal section of the tube. A pair of cylindrical and spherical lenses was used to transform the circular beam into a planar light sheet with dimensions of 20 mm wide by 0.5-mm thick. The horizontal light sheet coming from the laser was deviated by a 45 mirror so that the light entered vertically through the bottom wall of the visualization box and passed through the FEP pipe, illuminating a longitudinal section of the air-water flow inside the pipe. Filt er Laser sheet Camera Fig. 1 Schematic view of the test section. Optical setup for longitudinal film visualization. Images of the lower portion of the liquid film were captured using an IDT Motion Pro X3 camera operating from 250 to 3000 frames per second at a spatial resolution of 512 x 512 pixels. The camera was mounted orthogonally to the light sheet plane. Nikkor lenses with focal distances of 60 and 105 mm equipped with spacer rings were employed, respectively, for the film thickness and wave speed measurements. A TSI synchronizer was employed to synchronize laser firing and image capture. At the spatial resolution employed, the camera memory allowed for 52 seconds of image capture at 250 frames per second and for 4 seconds at 3000 frames per second. A high pass optical filter with a cutoff wave length of 560 nm was installed in front of the camera lens to block the 527- nm green laser light scattered by the air-water interfaces. With the filter installed, the camera only registered the 610-nm fluorescence light emitted by the Rhodamine dissolved in the water. Pixel calibration of the images was obtained by using a target inserted into the test pipe through its exit section, after the removal of the return pipe connection. The target was machined from a 1-m-long brass cylinder, forming a plane section passing through the cylinder diameter. On this plane section, a grid of regularly spaced vertical and horizontal lines was inscribed forming the calibration target. After the target plane was aligned with the vertical laser light sheet, the test pipe was filled with the same water and Rhodamine solution used in the tests. An image was then captured by the camera and the pixel calibration calculated by measurements made with the image acquisition software and the knowledge of the actual grid spacing. It should be mentioned that measurements made in the image at regions close to the lower pipe wall, at the pipe center line and close to the upper pipe wall, all presented - 3 -

4 the same pixel calibration value, indicating that no appreciable optical distortions were present. Cross sectional views of the liquid film were obtained by an optical setup employing two high frame rate cameras mounted at an angle, as indicated in Fig. 2. In this case, the light sheet was rotated by 90 so as to illuminate a cross section of the pipe. Two IDT Motion Pro X3 cameras were mounted at an angle of 45, imaging the pipe cross section through the two 45-degree-inclined windows provided at the visualization box that surrounded the test tube. Each camera was mounted on a support that permitted that the camera body was rotated in relation to the lens axis. This setup allowed the attainment of the Scheimpflug condition. When this condition is attained, the whole image is focused, even though the camera is viewing the pipe cross section at an angle (Raffel et al., 2007). Fig. 2 Top view of optical setup for cross section liquid film visualization. Calibration target, target as imaged by left and right cameras and after application of distortion procedure and joining operation. Image distortion due to the side camera viewing was corrected by a specially written program that used images of a cylindrical target captured by the left and right cameras. The target, shown in Fig. 2, was machined from 1-m long cylindrical brass piece, and was introduced in the test tube through its exit section, in the same fashion as described before for the side view experiments. A grid of regularly-spaced dots was machined on the target face. As part of the calibration procedure, the target was inserted into the FEP test tube and had its face aligned with the laser light sheet plane. The test tube was filled with the Rhodamine-water solution and one image of the target was captured with each camera and input to the distortion routine developed. The routine provided a distortion calibration polynomial for each camera that was later applied to each flow image captured. The pixel calibration value was also provided by the calibration procedure. Figure 2 shows the target images as captured by the left and right cameras and after the application of the distortion procedure and joining operation that will be described in the next section. It is relevant to mention that the calibration conditions described using a pipe completely filled with the water-rhodamine solution are distinct from those at flow conditions when a liquid film is flowing at the wall and there is an air core carrying liquid droplets. Due to this difference in optical paths, the image obtained from one camera can not be used to image the liquid film on the opposite pipe wall. As will be commented shortly, the right part of the liquid film was captured by the right camera, while the left part of the film was captured by the left camera. With this separation, it is guaranteed that the calibration polynomial is always applied within a liquid layer with no air layers in the optical path. 3. Image Processing The image processing procedures employed in extracting the film thickness versus time information from the set of captured images will now be outlined. Longitudinal images of lower liquid film. After a sequence of images of the lower liquid film - 4 -

5 for a particular combination of water and air flow rates was captured and stored, the first step in the image processing procedure was to specify the location and width of the two probes in the image where the thickness of the liquid film would be measured. Figure 3 was prepared to aid in the definition of the measuring probes in each image frame. l s1 l s2 d s Fig. 3 Schematic view of the probes employed for measuring the time-varying liquid film thickness. The figure is a pictorial view of a captured image frame where the lower and upper tube walls can be seen. The grid in the figure background corresponds to the camera pixels, while the curve represents an idealized instantaneous liquid film geometry captured by the camera. The two vertical strips marked in the figure are the selected probe regions where the film thicknesses are to be measured. The probes are spaced by a distance d s with widths given by l s1 and l s2. In the figure, the thicknesses measured in probes 1 and 2, corresponding to the time of capture t, are h 1 (t) and h 2 (t). The distance between the probes influenced the uncertainty on the wave speed measurements, while the probe width determined the level of spatial averaging to be applied to the thickness data. In order to keep the experimental uncertainty levels within acceptable limits, the experiments for measuring film thickness and wave speed were conducted separately using different pixel resolutions and acquisition frequencies. For the film thickness measurement experiments the optical system used a 105-mm lens, which produced a pixel resolution of 20µm/pixel. The probe width employed was equal to 50 pixels, which is equivalent to 1 mm in the flow domain. For the wave speed measurements, a 60-mm lens was employed producing a resolution of 55 µm/pixel. The probe spacing selected was equal to 150 pixels, which is equivalent to 8.25mm in the flow domain. In these experiments the probe widths were equal to 37 pixels or 2 mm. Prior to measuring the film thickness data, the images were processed with the objective of enhancing contrast by means of a histogram equalization procedure based on a sigmoid function that transforms the grey levels above and below a pre-determined value to, respectively, the maximum and minimum pixel grey values available in the camera sensor. The resulting image grey level histogram displays higher number of pixels concentrated on the extreme high and low values, with fewer number of pixels displaying intermediate values. This type of image histogram facilitates the determination of a threshold for the binarization operation that follows. A distinct feature of the histogram equalization procedure employed in the present study was its application to each individual image column, rather than to the whole image. Figure 4 presents a sample of a typical original instantaneous liquid film image captured by the camera. Below the image is the corresponding histogram. As can be verified, the number of pixels with values above 0.7 is negligible, with the majority of pixels concentrated in the zero-to- 0.6 range. Figure 4 shows the effect of the column-based histogram equalization procedure applied over the original image. The image contrast has been significantly enhanced and a binarization operation can be easily implemented. Figure 4(c) presents the image resulting from the binarization operation using an appropriate - 5 -

6 threshold value. The liquid film thickness at a determined axial position in the binary image can be easily determined by counting the number of white pixels until the first black pixel is found (interface), once the position of the lower wall is input to the program. In Fig. 4(d), as a verification procedure, a red line corresponding to the measured film thicknesses is overlaid on the original image (Fig. 4). The agreement obtained is considered excellent. The white patches over the liquid film, but not connected to it, are images of fluid out of the illumination plane, and we were not computed in the film thickness measurement. It should be mentioned that the image processing procedures just described were not applied on the entire images, as suggested by the examples shown in Figure 4. Rather, in order to save processing time, the image processing procedures were applied only on the regions defined as probes 1 and 2. (c) (d) Number of Occurrence Number of Occurrence Gray Level Gray Level Fig. 4 Original liquid film image and histogram. Image processed by column-based histogram equalization procedure and histogram. (c) Binary image. (d) Measured liquid film ovelaid on original image Cross section images of liquid film. The processing procedure applied to the liquid film images captured by the right and left cameras was initiated by applying a histogram equalization procedure to the image pair. This was necessary since the two images presented different grey level distributions due to differences in illumination. Here, contrary to what was previously described for the longitudinal film images, a global histogram procedure was applied to the images. Following, the distortion calibration polynomials that were previously determined with the aid of the calibration target were applied to each image. The resulting distorted left and right images can be seen in Fig. 5. Figure 5 presents the original left and right images, before distortion. Next, the two images were joined to form a complete instantaneous image of the film cross section. In order to assure a perfect matching of the left and right images, the position of the center of the calibration target as viewed by each camera was recorded during the calibration operation. These positions were used as a guide to match the images. Figure 5(c) presents the result of the image joining operation. A circular black mask was applied to the exterior of the pipe to trim ghost images resulting from film images out of the laser sheet and viewed by the cameras through the transparent tubes walls. This was merely a cosmetic operation with no implication on the film amplitude measurements. Figure 5(d) presents the joined images with the overlaid mask. The images were binarized and the film amplitude measurement was performed with the same procedure used for the longitudinal measurement already described. The amplitude measurements in the cross stream images were always made at the 0 position (bottom part of the film). Film amplitudes at other circumferential positions were obtained by rotating the cross section image to the 0 position by applying a rotation transformation to the image

7 4. Wave Characteristic Measurements Quantitative information on the liquid film wave behavior was extracted from the thickness versus time data measured from the longitudinal and cross section time resolved images of the liquid film. As already mentioned, liquid film thickness versus time and wave velocity data were obtained from separate experiments employing different pixel calibration values (different optical magnification) and different acquisition frequencies. For the amplitude versus time data, an acquisition frequency of 250 Hz was employed. The record was limited to 52 s by the camera memory. Time-averaged liquid film thickness was obtained from the amplitude data by averaging the amplitude data over the record length. This value includes the contributions of large amplitude waves, as opposed to the data reported by Schubring & Shedd (2009) that only considers the base film thickness variation. Root-mean-square (RMS) values of the film thickness data were also calculated to aid in the flow characterization. Power spectra densities (PSD) of the film thickness time records were calculated employing 256 Hamming windows to filter the results (Harris, 1978) that would, otherwise, be too noisy due to the camera memory limitation. (c) (d) Fig. 5 Liquid film images as captured by the left and right cameras. Distorted left and right images. (c) Joined images. (d) Joined images with overlaid mask. The film thickness data obtained by the longitudinal camera viewing, gives a 512 x 512 pixel resolution which was sufficient to yield acceptable accuracy in the thickness measurements, since only the region ranging from the bottom wall to the pipe centerline was imaged by the camera. For this pixel resolution the camera memory allowed the record length of images, which at 250 Hz, translated to 52 s of recording time. In the case of the stereoscopic cross section data however, each camera images half of the tube cross section, which requires the use of the 1024 x 1280 maximum pixel resolution offered by the camera in order to guarantee an acceptable accuracy in the measurements. With this pixel resolution, the camera memory allows capturing 6550 images, which correspond to a maximum recording time of 26 s at 250 Hz acquisition frequency. Due to the shorter record length, the PSD for the stereoscopic cross section data were calculated employing 128 Hamming windows. Cross correlation function (CCF) of the time-dependent liquid film thickness records measured at the locations of probe 1 and 2 were calculated to estimate the wave velocity (Bendat & Piersol, - 7 -

8 1971). The wave velocity was obtained by dividing the distance between probes 1 and 2, d s, by the time corresponding to the cross-correlation peak. Preliminary tests conducted indicated that the camera acquisition frequency should be increased to 3000 Hz so that the cross correlation results provided acceptable accuracy for the wave velocity data. The statistical calculations just described were implemented using routines available in the Matlab software. 5. Results and Discussion An experimental program was conducted with the objective of validating the optical technique developed. The experiments covered the ranges of water flow rates from to kg/s and air flow rates from to kg/s. These flow rates correspond to liquid superficial velocities of to m/s and air superficial velocities of 20 to 34 m/s. These operational conditions were chosen so as to allow comparison with results available in the literature. According to Taitel & Dukler (1976) flow map, these flow conditions are all within the annular flow regime region. All results to be presented were obtained with the 15-mm internal diameter pipe described in the experiments section, operating at atmospheric pressure level. Longitudinal and cross section visualizations. Since the technique developed is based on the digital processing of time resolved images of the liquid film, high quality instantaneous images of longitudinal and cross section views of the liquid film were registered and available for analysis. An examination of these images provides valuable visualizations of the dynamics of the film. Although in the present paper only the probe regions defined in each one of the images were analyzed for quantitative data extraction, samples of the complete longitudinal and cross stream images are displayed in Figs. 6 and 7. In the case of the longitudinal images (Fig. 6) the sample presented is part of a set of images of the film captured at the bottom of the tube. The images shown in the figure were captured at 3000 frames per second and display the passage of a liquid wave. Also seen in the figure are two marks, one red one green, at the air-water interface. These are the film thicknesses measured at the two probe locations by the image processing procedures. These marks were overlaid on the original images to serve as a visual verification of the accuracy of the image processing procedures employed. Fig. 6 Time-resolved longitudinal views of the liquid film at the bottom of the tube. Figure 7 presents a sample of cross section views of the liquid film around the tube perimeter captured at 2000 frames per second. The image sequence selected presents the passage of a large disturbance wave formed under the conditions of superficial air and liquid velocities of, respectively, 20 m/s and m/s. Visual analysis of a slow motion sequence of the images allows the observation of the circumferential motion of the liquid film climbing on the tube wall. Time-resolved liquid film thickness. Figure 8 presents a sample of time records of the film thickness at the bottom of the pipe obtained by analyzing the captured longitudinal image sequences by employing the image processing procedures developed. The results of Fig. 8 correspond to the film thickness measured at the location of probe 1 defined in Fig. 6. Figure 8 presents the film thickness for a constant liquid superficial velocity and for different values of the air superficial - 8 -

9 velocity. The presence of disturbance waves and ripples can be identified in the records presented. The results also show the decreasing trend of the film thickness and regularization of the large waves as the superficial gas velocity increases, as pointed out by several authors (Jayanti et al., 1990 and Paras & Karabelas, 1991). These results found in the literature were obtained employing electrical probes. Also in Fig. 8 are the space-time plots of the film thickness measured at the complete width of the laser sheet (approximately 20 mm) for a period of 0.05 s. These figures convey a clear image of the passage of disturbance waves and ripples. Fig. 7 Time-resolved images of cross section views of the liquid film for u sg =20 m/s and u sl =0.140 m/s h (mm) h (mm) Time (s) Time (s) (c) (d) Fig. 8 Time records of liquid film thickness at the bottom of the pipe obtained by analysis of longitudinal images. Data for u sl =0.112 m/s and superficial gas velocities equal to u sg =20 m/s and u sg =34 m/s. (c) and (d) Time space evolution of disturbance wave with ripples on it. Time-averaged and RMS film thickness. Average film thicknesses at the bottom of the pipe obtained from the measured time records for all operating conditions investigated are presented in Fig. 9. The results indicate that the average liquid film at the bottom of the tube is a decreasing function of the superficial gas velocity, depending weakly on the superficial liquid velocity for the range of flow rates investigated. These trends agree with previous results available in the literature (Jayanti et al., 1990 and Paras & Karabelas, 1991). RMS values of the film thickness at the bottom of the tube can also be extracted from the timeresolved thickness data obtained. In Fig. 9, the ratio of the RMS thickness data to the time

10 averaged thickness, is plotted as a function of the superficial gas velocity, for different values of the superficial liquid velocity. This ratio is a measure of the intensity of the film thickness fluctuation, and reaches a peak of 67% for the conditions corresponding to the lowest gas velocity (20 m/s) and intermediate liquid superficial velocity (0.084 m/s). These results are in agreement with the work of Paras & Karabelas (1991) that employed resistive probes h (mm) Usg (m/ s) h RM S / h Usl (m/ s) Usl (m/ s) Usg (m/ s) Fig. 9 Time-averaged liquid film thickness at the bottom of the tube. Ratio of the RMS liquid film thickness to the time-averaged thickness. Wave velocity results. Wave velocity measurements were obtained by cross correlating the time-resolved thickness data measured at the location of probes 1 and 2. A typical cross correlation function is presented in Fig. 10, where the abscissa indicates the number of image frames. The corresponding time lag can be obtained by dividing the number of frames by the camera acquisition frequency. In all tested cases, a 3000 Hz acquisition frame was utilized. In the example of Figure 13, the correlation peak is found at 4.7 frames, which corresponds to a time interval of 1.57 ms. The wave speed is obtained by dividing the probe distance, ds = 8.22 mm, by this time interval yielding a wave speed of 5.24 m/s. Figure 10 presents the wave velocity values obtained for all the experiments conducted. In this figure, measured wave velocities are plotted as a function of the superficial gas and liquid velocities. Not only was the correct dependence of the wave speed with liquid and gas superficial velocities captured by the optical technique developed, but also the numerical values obtained are in good agreement with the results from Schubring & Shedd (2008) and Fukano & Ousaka (1989) CCF v (m/ s) Usg (m/ s) Frames Usl (m/ s) Fig. 10 Cross correlation function of film thickness records measured at probe 1 and 2 for u sg =34 m/s and u sl =0.112m/s. Wave velocity measured at the bottom of the tube. Power spectra density (PSD) of thickness time records. Figures 11 and present PSD plots of the time-resolved thickness data for the flow conditions indicated. These PSD results were selected among all the flow conditions tested as an example to convey the capability of the optical technique implemented to extract spectral information from the film thickness data. The presentation of Fig. 11 conveys the influence of the gas velocity on the PSD of the film thickness

11 time data for two values of the liquid superficial velocity, namely, and m/s. The influence of the gas superficial velocity is significant for both liquid flow values. A clear increase in the dominant frequency with the superficial gas velocity is observed for both superficial liquid velocities, a trend also reported in the literature (Jayanti et al., 1990 and Paras & Karabelas, 1991) PSD (db/ Hz) Usg (m/ s) PSD (db/ Hz) 8 6 Usg (m/ s) Frequency (Hz) Frequency (Hz) Fig. 11 PSD of film thickness time variation for different gas superficial velocities. u sl =0.084 m/s. u sl =0.112 m/s. Comparison between longitudinal and cross section measurements. Figure 12 was prepared to present a comparison on the longitudinal measurements with those obtained by the cross sectional measurements. In the figure the PSD obtained by the longitudinal and stereoscopic techniques for u sg =20 m/s and u sl =0.140 m/s are compared for the bottom position of the tube. The agreement obtained is considered excellent. For the same flow conditions of Fig. 14, the average film thickness obtained by the longitudinal and stereoscopic techniques were, respectively, 0.88 and 0.87 mm what attests for the potential of the stereoscopic technique. As a sample of ongoing work, Figure 12 presents an instantaneous image of the cross section of the flow, where the film thickness was measured by the technique developed at every 5 degrees around the tube circumference. A line was then fitted to the points measured providing an instantaneous measurement of the complete liquid film thickness PSD (db/ Hz) Stereo Longitudinal Frequency (Hz) Fig. 12 Comparison of PSD of film thickness data at the bottom of the tube obtained by the longitudinal and stereoscopic optical techniques for u sg =20 m/s and u sl =0.140 m/s. Film thickness around the tube perimeter. 6. Conclusions The present paper presented an optical technique developed for measuring the statistical and spectral properties of time-varying liquid film thicknesses in horizontal, air-water annular flow. The technique proposed builds on existing visualization techniques described in the literature. It uses tube material with index of refraction that nearly matches that of water in order to permit the visualization of thin liquid films adjacent to the tube wall. Laser induced fluorescence was employed to separate the intense light reflected from the air-water interfaces and allow the recording of the desired liquid film images. Recording of the images was conducted with high frame rate cameras what produced time-resolved data with good spatial resolution. The quality of

12 the film images obtained allowed the visualization of the wave behavior of the liquid film. Two versions of the optical technique were implemented. In one technique, a longitudinal section of the film defined by a pulsed laser light sheet was imaged with a high frame rate camera synchronized with laser firing. The second technique employed two high frame rate cameras in a stereoscopic setup to yield an instantaneous view of the complete cross section of the liquid film. Specially developed image processing routines were applied to improve image contrast, to calibrate the images, and to correct for the distorted views associated with the stereoscopic setup. Quantitative information on the statistical properties of the liquid film was extracted from the digital images for both, the longitudinal and stereoscopic setups. The processed results included time-resolved film thickness, time-averaged and RMS thickness values, wave velocities and power spectra density of the thickness data. Good agreement was obtained with data from the literature, which served to validate the technique. 7 Acknowledgements The authors gratefully acknowledge the support awarded to this research by Petrobras R&D Center. Our gratitude is also extended to the Brazilian Research Council, CNPq. 8 References Bendat, J.S. & Piersol, A.G. Random data: Analysis and measurement procedures. Wiley- Interscience, New York, Butterworth, D. & Pulling, D.J. A visual study of mechanisms in horizontal annular air-water flow. Atomic Energy Research Establishment, M-2556, Berkshire, Fukano, T. & Ousaka, A. prediction of the circumferential distribution of film thickness in horizontal and near-horizontal gas-liquid annular flows. Int. Journal of Multiphase Flow. Vol. 15, Harris, F.J. On the use of windows for harmonic analysis with the discrete fourier transform. Proceedings of the IEEE. Vol. 66, Hewitt, G.F., Jayanti, S. & Hope, C.B. Structure of thin liquid films in gas-liquid horizontal flow. Int. Journal of Multiphase Flow. Vol. 16, Jayanti, S., Hewitt, G.F. & White, S.P. Time-dependent behaviour of the liquid film in horizontal annular flow. Int. Journal of Multiphase Flow. Vol. 16, Koplin C.R. Local Liquid velocity measurements in horizontal, annular two-phase flow. PhD Thesis. University of Wisconsin-Madison Paras, S.V. & Karabelas, A.J. Properties of the liquid layer in horizontal annular flow. International Journal of Multiphase Flow. Vol. 17, Raffel M., Willert C.E., Wereley S.T. & Kompenhans J. Particle image velocimetry A practical guide. 2 nd Ed. Springer Berlin Heidelberg, New York, Rodríguez, D.J. & Shedd, T.A. Cross-sectional imaging of the film in horizontal two-phase annular flow. ASME Heat Transfer/Fluids Eng. Summer Conference, Charlotte, NC, USA, Schubring D. & Shedd T.A. Critical friction factor modeling of horizontal annular base film thickness. Int. Journal of Multiphase Flow. Vol. 35, Schubring, D. & Shedd, T.A. Wave behavior in horizontal annular air-water flow. Int. Journal of Multiphase Flow. Vol. 34, Shedd, T. A. & Newell, T.A. Automated optical liquid film thickness measurement method. Review of Scientific Instruments. Vol. 69, Sutharshan, B., Kawaji, M. & Ousaka, A. Measurement of circumferential and axial liquid film velocities in horizontal annular flow. Int. Journal of Multiphase Flow. Vol. 21, Taitel, Y. & Dukler, A.E. A model for predicting flow regime transitions in horizontal and near horizontal gas liquid flow. AIChE Journal. Vol. 22, Taylor, N.S.H. & Nedderman, R.M. The coalescence of disturbance waves in annular two phase flow. Chemical Engineering Science. Vol. 23, Vassalo, P. Near wall structure in vertical air-water annular flows. Int. Journal of Multiphase Flow. Vol