Loading...
Please wait, while we are loading the content...
Similar Documents
PIV measurements of waves and turbulence in stratified horizontal two-phase pipe flow
| Content Provider | Semantic Scholar |
|---|---|
| Author | Birvalski, M. Tummers, Matthias Delfos, René Henkes, R. A. W. M. |
| Copyright Year | 2013 |
| Abstract | The technique for obtaining detailed velocity fields in a wavy liquid layer in stratified air/water pipe flow is described in this paper. By combining Particle Image Velocimetry (PIV) with an interface detection technique, the velocity field is resolved in the whole liquid layer. Furthermore, since the shape of the interface is resolved at each time instance, this information is used to conditionally average the velocity field according to the wave phase, which results in phaseresolved velocity profiles. These velocities are then used to separate the wave-induced motion from the turbulenceinduced motion in the liquid layer. In this way, the turbulent wavy regime is analysed. The results of the measurements are compared to the theory of waves and turbulence. INTRODUCTION Stratified flow of gas above a liquid layer through closed conduits such as channels or pipes is encountered in many industrial applications. The liquid layer can be either laminar or turbulent, and the interface can be either smooth or wavy. The knowledge of the velocity field of the liquid layer, and the characteristics of the interface, are important both from a scientific as well as from an engineering point of view. Studies that obtained the velocity profiles in the stratified liquid layer were initially performed in rectangular channels [1-4], and later also in pipes [5-7]. These studies showed that when the interfacial drag is absent, or is very low due to the low air velocity, the liquid velocity profile resembles the profile of a Couette-type flow. The interface in this case acts as a moving wall, dampening the turbulent fluctuations in the surface-normal direction, and slightly enhancing the fluctuations in the surface-parallel direction. When the shear is larger, however, and waves are present on the interface, the velocity profile attains a characteristic S-shape. The turbulent fluctuations show peaks close to the wall due to shearinduced turbulence, but also close to the interface due to fluctuations introduced by the wavy motion. To distinguish between the velocity fluctuations due to waves from those due to turbulence, the knowledge of the interfacial profile is required. Several studies [8-10] report on techniques that simultaneously acquire the velocity field (by using PIV) and the interfacial profile in flumes or wind-wave tanks. In these studies, two cameras are used: one for capturing PIV images and another that records the interface shape. The PIV camera is positioned below the interface, and angled upwards, while the Profile camera is positioned above the interface, and angled downwards. The interface shape is extracted from the Profile camera images by using an edge detection algorithm, since the illuminated particles in the liquid layer form a sharp edge against the dark background. When both the velocity field and the interfacial profile are known, a phase-averaging procedure can be applied to decompose the velocity field into the time-averaged mean field, the wave-induced motion and the turbulence-induced motion, see e.g. [11, 12]. Thus far, the only study that reports on using PIV in stratified gas/liquid pipe flow is the one by Vestøl and Melaaen [7]. They did not, however, record the interface shape, and therefore the velocity field could not be decomposed. In the present paper, the procedure that combines PIV with interface detection is applied to stratified gas/liquid pipe flow. In this way, the influence of the two main governing phenomena – namely the waves and the turbulence – on the resulting velocity field can be studied separately. EXPERIMENTAL SETUP A straight 10.3 m long, 50 mm ID pipe, made of transparent PMMA and high-quality glass (for the PIV test section) was positioned horizontally (see Figure 1, left). The laboratory air was circulated using a fan, and the flowrate was measured with an Instromet turbine flow meter. The water was circulated with a pump, and the flowrate was measured with a Rheonik Coriolis mass flow meter. Both fluids were at atmospheric pressure and temperature. An inlet section was positioned at the entrance which had a Y shape, with water entering horizontally, and air entering at 30 degrees from above. A development length of 150 pipe diameters (7.5 m) was provided before the PIV measurement section. This section, made of Schott Duran glass, included a box-shaped extension (made of regular glass), which corrected for most of the distortions produced by the curved wall of the pipe, and facilitated focusing inside the liquid layer. Figure 1 Experimental setup. Schematic view (left) and a cross-section through the PIV test section (right). The PIV system consisted of a Continuum Minilite Nd:YAG laser (double-shot, 532 nm wavelength, approximately 1.2 mJ/pulse) firing at 2 Hz, a double-frame CCD camera (PCO Sensicam qe, 1376x1040 pixels, 12-bit grayscale) equipped with a 100 mm Tokina lens, and an ILA PIV-Sync synchronizer unit. Like in [8-10], the camera was positioned below the level of the interface and it was set at an angle of approximately 16 degrees, looking upwards underneath the waves (Figure 1, right). This position was necessary to image the whole liquid layer, including the interface, and to avoid any blocking of the view due to waves. The profile capturing was performed with another CCD camera (PCO Sensicam, 1280x1024 pixels, 12-bit grayscale), but positioned above the interface, looking at approximately 7 degrees downwards. The lenses of both cameras were fitted with long-pass optical filters, made from Hoya O56 coloured glass. These filters blocked the green laser light (reflected from the pipe wall and the interface) but transmitted the higher wavelength light emitted by the seeding particles present in the liquid. The material for seeding is a fluorescent pigment from a commercially available acrylic paint, produced by Lefranc & Bourgeois (Flashe Vinyl, fluorescent light orange colour). The pigment particles in this paint have a range of sizes of roughly 1 to 30 μm, with particle agglomerates reaching ~50 μm. Before using them for seeding, the smallest and the largest particles needed to be discarded. This was done by performing two steps of sedimentation. In the first step, the dispersed paint was left to settle over several hours. The sediment, which contained the largest particles, was discarded and the suspension left to settle over a longer period (typically 48 h). After this, the suspension containing the smallest particles was discarded and the sediment was used for seeding. The light from the laser passed through a set of lenses which shaped it into an approximately 0.5 mm thick and 35 mm wide light sheet in the vertical streamwise plane in the middle of the pipe. The cameras were positioned approximately 40 cm away from the pipe centreline, and their field of view was approximately 25x19 mm. A grid of dots (0.5 mm diameter, spaced 1 mm apart), was positioned vertically in the centre plane of the pipe in order to capture the calibration images. An image for the profile camera was taken with the pipe filled with air, and an image for the PIV camera was taken with the pipe filled with water. These images were used to dewarp and scale the captured images, which resulted in corrected images from both cameras that have a common coordinate system. The PIV images were processed using LaVision’s Davis 8 software, by applying a multi-pass, multi-grid, adaptive cross-correlation procedure with the final interrogation area size of 16x16 pixels. The velocities were overlapped by 50%, which resulted in a grid of vectors spaced 0.148 mm apart (the resolution is ~0.0185 mm/pixel). The outliers were detected and eliminated using standard procedures. The final processing step was the application of a mask, generated using the detected interface profile, which sets to zero all the velocity vectors that are outside the liquid layer, i.e. above the air/water interface. PROFILE DETECTION A profile detection technique is developed and presented through a series of images shown in Figure 2. Similar algorithms were developed previously by e.g. [8-10] and applied to wave tanks and flumes. We developed the technique further to cope with the small pipe geometry and with the reflections that arise due to the curved walls surrounding the liquid layer. Figure 2a is a part (700 pixels wide) of an image captured by the Profile camera. The bright area in the lower part consists of images of particles in the light sheet, refracted through the interface. The interface which needs to be detected is seen as a sharp edge in the central part of the image. Besides the interface, two bright spots are seen that lie beyond the edge of the light sheet. These spots are due to light that has been refracted by the interface and then reflected from the top wall of the pipe back onto the liquid layer (see Figure 1, right). This light illuminates the fluid away from the light sheet and, although being much less intense than the original sheet, can create a strong signal if it encounters a large tracer particle on or near the interface. The algorithm, implemented using the Image Processing Toolbox of Matlab (Matlab R2012a, The MathWorks Inc., USA), incorporates steps that correct errors in the detection due to these reflections. Figure 2 Step-by-step processing of a sample Profile image. a) raw image (detail), b) image after Sobel and Wiener filtering, c) image after thresholding, d) binary representation of the interface, e) interface after opening and closing, f) final profile after smoothing, g) final profile overlaid on the corresponding PIV image (detail). In the first step, a 3x3 averaging filter is applied to the original image, followed by a vertically oriented Sobel operator, 3x3 pixels in size, which accentuates the edges of the light sheet. Next, a Wiener filter (11x11 pixels) is applied which smoothens the previous result. The absolute value at all pixel positions is calculated and the resulting |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://repository.tudelft.nl/islandora/object/uuid:0ac3cbb2-4259-46cd-8587-8b5ed4566581/datastream/OBJ/download |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
