International Journal of Multiphase Flow

Transcription

1 International Journal of Multiphase Flow 67 (214) Contents lists available at ScienceDirect International Journal of Multiphase Flow journal homepage: Developing air water flow downstream of a vertical 18 return bend q Pedro M. de Oliveira, Eduardo Strle, Jader R. Barbosa Jr. POLO Research Laboratories for Emerging Technologies in Cooling and Thermophysics, Department of Mechanical Engineering, Federal University of Santa Catarina (UFSC), Florianópolis, SC 8849, Brazil article info abstract Article history: Received 31 January 214 Received in revised form 6 May 214 Accepted 7 May 214 Available online 22 May 214 Keywords: Two-phase flow Return bend Gas holdup Developing length Frictional pressure drop Flow visualization The importance of developing two-phase flow downstream of return bends has been brought about by several authors wishing to quantify irreversible losses in singularities. In a recent work, a method for evaluating such losses in vertical return bends was proposed based on gas holdup and pressure drop measurements in the bend. In the present study, the nature of the developing flow upstream and downstream of a vertical 18 tube bend (curvature ratio of 8.7) is investigated by means of gas holdup and pressure drop measurements at several positions along the 26.4-mm ID tubes connected to the bend. The effect of the bend on the flow characteristics was evaluated for the plug, slug and annular flow regimes in the upward and downward flow directions. The results showed that, although the variations in gas holdup seemed to extend beyond the measuring limits, significant pressure changes were limited to 4 pipe diameters upstream and 6 diameters downstream of the bend. Ó 214 Elsevier Ltd. All rights reserved. 1. Introduction Return bends are widely encountered in industrial equipment and piping systems. Two-phase flows in vertical bends are affected by centrifugal and gravitational forces that modify the configuration of the phase interfaces (flow regimes) and change the flow parameters, e.g., pressure gradient and gas holdup, downstream of the bend. These changes vanish after a certain distance downstream commonly referred to as the flow developing (or recovery) length. In a recent work (de Oliveira and Barbosa, 214), the gravitational and acceleration components of the total pressure drop in gas liquid flows in vertical return bends were determined experimentally for a range of flow conditions. In some cases (e.g., downward flows at low mixture velocities), the static head was found to be the most critical term, corresponding to about 5 times the frictional pressure drop. The reversible pressure change associated with the acceleration of the flow also accounted for a significant q To Geoff Hewitt on his 8th birthday. I treasure every minute I spent with GFH since we first met in Florianópolis in I could not imagine then how that brief moment would change my life as I knew it. To this day, every discussion with him is still as fresh and rewarding as when I first showed him my calculations on the droplet concentration imbalance and how these explained the heat transfer coefficient decrease in annular flow of mixtures. His enthusiastic reply ( There is nothing like it, Jader! ) synthesized the passion for science and the pleasure of discovery that I strive to pass on to my students every day. Happy birthday, Geoff! Corresponding author. Tel./fax: address: jrb@polo.ufsc.br (J.R. Barbosa Jr.). part of the pressure drop, reaching twice the value of the frictional component in some situations. Despite their non-negligible contribution to the total pressure drop, the reversible components of the two-phase pressure gradient are commonly neglected in experimental analyses where a simultaneous assessment of the phase holdup is not available. A simplified way of estimating the frictional pressure drop in a bend is to measure the pressure drop in a segment that includes the bend and a straight-pipe section downstream of it. In this case, the frictional component of the pressure drop along the whole segment is taken as an average value, assuming that the change in flow momentum has diminished and the flow is developed at the pressure tap furthest downstream (Usui et al., 198). The issue of flow development downstream of return bends has received some attention in the open literature, being first addressed by Traviss and Rohsenow (1971), who measured the pressure drop and condensation heat transfer coefficient of R-12 along a horizontal segment (4.4-m long, 8-mm ID) immediately following a vertical return bend. They concluded that the effect of a return bend on the downstream pressure drop was negligible when averaged over a length of 9 diameters, and that the pressure gradient did not deviate more than 1% from the fully-developed pressure gradient. In their work, two bends made of glass with curvature ratios (defined as the ratio of the bend radius to the pipe radius) of 3.2 and 6.4 were used to allow flow visualization with a high-frequency light source. The flow pattern was found to readjust very rapidly when disturbed by the presence of the bend; a fact that was verified years later by da Silva Lima and Thome /Ó 214 Elsevier Ltd. All rights reserved.

2 P.M. de Oliveira et al. / International Journal of Multiphase Flow 67 (214) (212) and De Kerpel et al. (213) with more sophisticated and precise methods. A high-speed cine analysis of two-phase air water flows in vertical return bends was presented by George (1971), who demonstrated that disturbance waves are somewhat destroyed as annular flow passes through the bend in upward flow. The highspeed film contains axial viewing sequences of the flow in the outlet of the bend, which helped to understand the influence of the bend on the phase distribution in annular flow. The original 16- mm film produced at Harwell was kindly made available to the present authors by Prof. G.F. Hewitt. It was recently converted to digital video for the occasion of this publication, and can be accessed from the following link: Usui et al. (198, 1981) carried out air water flow pressure drop experiments in 24-mm vertical return bends with curvature ratios of 11.3, 16.6, and In order to evaluate the frictional pressure drop contribution, the authors identified the need for a precise assessment of the static head in the return bend. Usui et al. (198) used a pair of quick-acting solenoid valves at the inlet and outlet of the bend, which were simultaneously closed so the liquid content in the bend could be measured and, from that, an average value of the gas holdup in the bend could be estimated. They observed that the average gas holdup in the bend was not significantly influenced by the centrifugal force in the upward flow direction and, therefore, presented a good agreement with the Smith (1969) correlation. The opposite was observed for downward flow, where the gas holdup values differed significantly from those in straight tubes. The local gas holdup was measured by Usui et al. (1981) at the outlet of the bend and at positions of approximately 66 diameters upstream and downstream of it using an electrolytic probe. Measurements in the plug flow regime showed that the local gas holdup at the outlet of the bend was significantly higher than upstream due to acceleration of the liquid phase at the bottom part of the bend. At the downstream position, the cross sectional profile of the gas holdup was very similar to the one upstream of the bend, suggesting the existence of developed flow. Besides measuring pressure drop and gas holdup in the bend, Usui et al. (1981) also conducted visual observations of the flow and presented details on the different phenomena observed, e.g., flow reversal and flooding in the upward direction, and backflow of bubbles and coalescence in the downward flow direction. Hoang and Davis (1984) conducted air water flow experiments in return bends connecting two 25.4-mm ID vertical tubes with curvature ratios of 4 and 6, in an inverted U configuration. Their study was limited to the bubbly flow regime, as only high liquid mass fluxes were used. Static pressure was measured downstream and upstream of the bend, and within the bend itself. For the latter measurements, the pressure taps were distributed every 3 from the inlet to the outlet, both on the inner and outer parts of the curve (i.e., concave and convex parts). Pressure losses as large as 2 times those in single-phase flow were observed and associated with the separation and remixing of the phases in the bend. The authors identified a developing length of 9 diameters downstream of the bend, after which the flow was considered to be well remixed. This value is 1 times lower than the one observed by Traviss and Rohsenow (1971), probably due to the very high liquid flow rates used in the more recent experiments. By comparing the angular pressure profiles with the high-speed film footage, it was verified that the onset of rotation and stratification (separation) of the flow occurred in and after the first half of the bend, respectively. Studies focusing on the influence of the bend on the flow regimes of an air water mixture were carried out by Wang and co-workers in horizontal return bends (Wang et al., 23, 24) and vertical return bends (Wang et al., 25, 28). In these works, the distribution of the phases was observed via still photography. Return bends of mm ID and curvature ratios of 3, 5 and 7 were used in the horizontal experiments, while a single bend geometry was used in the vertical experiments (6.9 mm ID and 2R=d ¼ 3, where R is the bend radius and d is the pipe diameter). Phenomena such as flow regime transition from stratified to annular flow were observed in horizontal bends, being more pronounced in the small curvature radii and large pipe systems. In the vertical bends, flow reversal and frozen slug flow were observed. In their experiments using R-134a and horizontally-oriented return bends (13.4 mm ID and curvature ratio of 9), da Silva Lima and Thome (21) observed changes in the pressure gradient as far as 141 diameters downstream of the bend. Later, da Silva Lima and Thome (212) conducted two-phase flow experiments with R-134a in horizontal and vertical return bends. Glass return bends (internal diameters of 8, 11 and 13 mm, and curvature ratios of approximately 3 and 5) were used to allow visualization of the flow with a high-speed camera. The recovery length downstream of the bend was evaluated qualitatively based on the visual characteristics of the flow, and was found to be larger for vertical bends, specially in upward flow. The authors observed a larger influence of the centrifugal force on the flow rather than the effect gravity, probably due to the small curvature ratios. Several flow phenomena observed by the authors were in agreement with the observations of Usui et al. (198, 1981) and Traviss and Rohsenow (1971), such as liquid segregation and droplet deposition on the outer part of the curve. Recently, De Kerpel et al. (213) developed a method for determining the downstream development length based on the measurement of the flow capacitance by a probe placed on the outer pipe wall. A clustering algorithm was used to group similar signals associated with specific flow regimes for developed flow conditions in straight tubes. Tests were performed with R-134a in an 8-mm ID return bend with a curvature ratio of Measurements of flow capacitance were taken at different positions downstream of the bend (from 2.5 to 31.5 diameters) and compared with those obtained for developed flow conditions. Flow disturbances were detected as far as 1 diameters downstream of the bend. Padilla et al. (213) used pressure drop data to determine the perturbation lengths downstream and upstream of the bend in downward flows of R-134a in vertical return bends (6.7 mm ID, curvature ratio of 7.46) for total mass fluxes of 57 and 95 kg/ m 2 s. The total bend pressure drop was also measured for R-1234yf and R-41A for downflow in bends of 7.9 and 1.85 mm ID and curvature ratios of 3.68 and 4.5, respectively. The gravitational term of the total bend pressure drop was estimated using the gas holdup given by the Steiner (1993) correlation for straight tubes. Perturbation lengths of the downward flow were determined as being 5 diameters upstream and 2 diameters downstream of the bend. The present work deals with the experimental characterization of the pressure gradient and gas holdup in air water two-phase flows upstream and downstream of a vertically-oriented 18 return bend. The bend connects two 5-m long, 26.4-mm ID horizontal tubes, and the flow can be set as upward or downward (i.e., entering from the top or bottom tube). The bend curvature ratio, 2R=d, is 8.7. Measurement of the static pressure and gas holdup downstream and upstream of the return bend allows the evaluation of the components of the pressure gradient due to friction, acceleration and gravity. Three flow regimes are investigated: plug, slug and annular flow, for a superficial liquid velocity of.2 m/s and superficial gas velocities of.4, 4 and 2 m/s, respectively. The phase distribution in the bend is investigated with a high-speed camera, illustrating several features of the two-phase flow in the bend in the upward and downward directions.

3 34 P.M. de Oliveira et al. / International Journal of Multiphase Flow 67 (214) Experimental work 2.1. Experimental setup The experimental apparatus (Fig. 1) used in this work was presented in detail by de Oliveira and Barbosa (214), so only its main features will be reproduced here. The setup contains two individual fluid flow lines (air and water) equipped with transparent inlet flow mixers (1) through which the two phases are introduced. The test section consists of a 18 bend connecting the two horizontal tubes (26.4-mm ID), which are made from borosilicate glass and are approximately 5 m long. The 26.4-mm ID return bend, also made of borosilicate glass, comprises the C-shaped section and two straight sections of approximately.1 m in length, which connect the bend to the 5-m long tubes. The air water mixture flows from either of the two mixers and can go upwards or downwards through the return bend, depending on the mixer used as the flow inlet. The curvature ratio of the bend employed in this study was 2R=d ¼ 8:7. The main limitation for the value of the curvature ratio is the size (i.e., external diameter) of the inlet flow mixers. Nevertheless, the smallest curvature ratio evaluated in this facility was 6.2 (de Oliveira and Barbosa, 214), which, although slightly large, is still within the same range of the curvature ratios investigated elsewhere. The influence of the bend on the straight segments upstream and downstream of it has been investigated through the distribution of flow parameters along the axis of the tube. Absolute pressure was measured at taps located 1.85 m upstream and downstream of the return bend. Pressure drops were measured between the inlet and outlet of the bend and between sections of.11,.22,.51, and 1.1 m in length along 1.85-m long segments in the upper and lower tubes (Fig. 2). The pressure taps are precision machined in the perspex unions that connect two adjacent pipe segments. The taps are located in the middle of the perspex union, at a 5 angle from the bottom of the union. Fig. 3 shows the dimensions of a perspex union, which contains two o-ring grooves on each side for improved sealing. The diameter of the pressure tap is.8 mm. The tubes are mounted flush on each side of an inner ring (1-mm wide, 26.4 mm in diameter). Differences smaller than.2 mm were measured between the internal pipe diameter and the inner ring of the perspex union due to losses in circularity arising from the glass tube manufacturing process. Gas holdup was measured with two non-intrusive capacitance sensors positioned at twelve different locations of the test section. Therefore, six combinations for the positions of the pair of sensors (illustrated by the lines connecting the a symbol in Fig. 1) were needed to cover the six measurement points along each horizontal tube, as shown in Fig. 2. Six independent experimental runs were needed to fully characterize each experimental condition (i.e., combination of gas and liquid superficial velocities). It should be noted from Fig. 2 that the origin of the main coordinate system (z = ) is the position of the pressure taps closest to the bend. The capacitance sensors consist of two electrodes mounted flush on the outer wall of the tube (Libert et al., 211; Libert, 213), as shown in Fig. 4. The axial length of both electrodes in each sensor is 32 mm. The circumpherential lengths of the transmitter and receptor electrodes are 45 and 32 mm, respectively. The corresponding arc angles (transmitter and receptor) are and The receptor electrode is smaller because of the two guard electrodes on each side. For the plug and slug flow tests, an experimental calibration based on the static stratified configuration was used (de Oliveira and Barbosa, 214). In his comparisons between the capacitance sensor measurements and the wire-mesh measurements, Libert (213) reports that the phase distribution in slug flow is quite similar to that of stratified flow over a significant portion of the slug unit. To some extent, these similarities in phase distribution were also verified in our work through visual observations of the flow (although the interface is visibly more disturbed, i.e., rougher, in slug flow). A second calibration curve was employed to increase the sensor accuracy in the annular flow regime, as depicted by a dashed line in Fig. 5. This calibration, performed by Libert (213), was based on finite-element method (FEM) simulations of the electrical admittance of the region between the electrodes (tube wall and gas liquid flow) for the annular and dispersed bubble flow regimes Fig. 1. Schematic diagram of the experimental apparatus. (1) Inlet flow mixers, (2) thermostatic bath, (3) centrifugal pump, (4) Coriolis mass flow meter, (5) particulate filter, (6) coalescent filter, (7) pressure regulator, (8) micrometric valve, (9) hot wire anemometer-based mass flow meter, (1) water feed.

4 P.M. de Oliveira et al. / International Journal of Multiphase Flow 67 (214) Fig. 2. Positions of the pressure taps and gas holdup sensors. The measurements were taken at symmetric locations between the upper and lower tubes. Fig. 3. Dimensions (in mm) of a perspex union and positions of the pressure taps and thermocouple wells. (Libert, 213). For annular flow, Libert (213) assumed a symmetric film (smooth), with no droplet entrainment. The thickness of the liquid film was varied so as to simulate different values of the gas holdup between 6% and 1%, with steps of 5%. Fig. 5 also shows the experimental calibration data and curve fits for both sensors, including an additional FEM-based calibration for the stratified flow regime provided by Libert (213), and shown here for comparison purposes only (dotted line). The standard uncertainty associated with the present gas holdup measurement technique was evaluated taking into account the approach by Libert (213) for each flow regime using its respective calibration curve. For plug and slug flows, the data were compared against reference values obtained with a wire-meshsensor (da Silva et al., 27). The standard uncertainty was calculated based on the maximum error between the measured and the reference values, being 8% for plug and slug flows. Libert (213) did not compare the results of the capacitance sensor with the wire-mesh sensor for the annular flow regime. However, when the annular flow calibration curve was used to predict the slug flow data ð j l ¼ :3 m/s, j g ¼ 2 m/s), the capacitance sensor gas holdup results were within 7.2% of the reference (i.e., wire-mesh) data (Libert, 213) Experimental procedure The initialization of the experimental apparatus consists of positioning the gas holdup probes on the test section and acquiring its calibration parameters, i.e., the voltage signals for single-phase flow of air and water. The gas holdup signal is normalized by the single-phase flow signal, and a particular calibration curve is applied (de Oliveira and Barbosa, 214). Additionally, any air Fig. 4. (a) Test section and gas holdup capacitance probes, 2R=d ¼ 8:7, (b) capacitance probes. bubbles inside the tubing connecting the pressure transducers to the test section are purged from the system. As the pump is switched on and the pressure regulator is opened, mass fluxes of air and water are set by adjusting the valves and the pump speed until the desired phase superficial velocities are reached. Tests were carried out for three flow regimes in both upward and downward directions, accounting for six distinct flow conditions. For each condition, the superficial liquid velocity, j l, at the inlet of the measuring section (marked as ref. in Fig. 1) was set as.2 m/s, and the superficial gas velocity, j g, was varied between

5 36 P.M. de Oliveira et al. / International Journal of Multiphase Flow 67 (214) α [-] FEM Ann V* [-] the nominal values of.4, 4 and 2 m/s, corresponding to the plug, slug and annular flow regimes, respectively. After the measurements were taken, the initialization procedure was executed again for the new positions of the gas holdup probes. In total, 36 independent experimental runs were performed, resulting from the pair of gas holdup probes being placed at six different positions along the top and bottom tubes (Fig. 2) for each of the six combinations of liquid and gas superficial velocities and flow direction. The variations in the gas and liquid superficial velocities between runs related to each experimental condition were taken into account in the calculation of the expanded uncertainties reported in the electronic annex (Supplementary data file). The procedure for calculating the uncertainties in the present experiments is identical to the one reported in a previous paper (de Oliveira and Barbosa, 214) Frictional pressure drop The frictional pressure drop in a straight tube segment, Dp f, was evaluated as follows, ; ð1þ Dp f ¼ Dp þ G q out q in where the second term is the accelerational component of the total (measured) pressure drop, Dp ¼ p out p in. In the above equation, G is the total mass flux, and the subscripts in and out stand for the inlet and outlet sections of the tube segment, respectively. The momentum density is given by, " # 1 q ð1 xþ2 ¼ q l ð1 aþ þ x2 ; ð2þ q g a Sensor 1 Sensor 2 Fit 1 Fit 2 Fig. 5. Calibration curves of the gas holdup sensors. FEM Strat. where a is the local gas holdup. In Eq. (2), x is the dynamic gas mass fraction, i.e., the ratio of the air and total mass flow rates. When evaluating the frictional pressure drop in the segment containing the return bend, the contribution of the static head to the total pressure drop must be accounted for. Thus, Dp f ;b ¼ Dp þ G qg2r; ð3þ q out q in where the sign of the gravitational term is positive for upward flow and negative for downward flow. R is the bend radius, g is the gravitational acceleration, and q is the average density of the mixture in the bend, which is taken as the arithmetic average of the mixture density at the inlet and outlet of the bend. The mixture density is given by, q ¼ q g a þ q l ð1 aþ; where the subscripts g and l denote the gas and liquid phases, respectively. The gradients of the pressure drop components were calculated as follows, rp f ¼ Dp f ; ð5þ z out z in and, rp a ¼ G2 1 z out z in q out ð4þ 1 q in ; ð6þ where z in and z out are the axial positions of the inlet and outlet sections of the tube segment. For consistency in the presentation of the data, the pressure gradient components calculated according to Eqs. (5) and (6) are associated with the positions in the middle of each tube segment. 3. Results and discussions In the absence of phase change in a horizontal tube, a gas liquid flow is said to be developed when the frictional pressure gradient and the gas holdup remain fairly unchanged along the tube (Traviss and Rohsenow, 1971; da Silva Lima and Thome, 212). In the present paper, observation of the behavior of these parameters as a function of the axial position was used to evaluate how the local disturbances associated with the bend are propagated in the upstream and downstream directions for the different flow regimes Gas holdup measurements Fig. 6 depicts the gas holdup distribution in upward and downward flow for the plug, slug, and annular flow regimes. The lower horizontal axis is the distance in relation to the reference position at the inlet of the gas holdup measuring section z ref see Fig. 2 and the upper horizontal axis is the distance in diameters to the inlet (negative values) and from the outlet (positive values) of the bend, with respect to the main axial coordinate, z. It should α [-] in out Upward Downward.2 Plug Slug Annular * Fig. 6. Gas holdup distribution in upward and downward flow for the plug, slug, and annular flow regimes.

6 P.M. de Oliveira et al. / International Journal of Multiphase Flow 67 (214) be noted that, in terms of z ref, the distance between the data points corresponding to the inlet and outlet of the bend is the actual length of the pipe bend (in meters). On the other hand, for the scale based on the number of pipe diameters, both the inlet and outlet sections of the bend (in and out on the top horizontal axis) correspond to z=d =. In this way, negative values of z=d are always associated with the flow upstream of the bend (and positive values with the flow downstream of it), irrespective of the flow orientation in the test section (upward or downward). Of all flow regimes, plug flow presented the largest variations of gas holdup, being visibly more affected by the bend. As the gas velocity increases and the flow regime changes to slug and annular, the influence of the bend on the gas holdup seems to decrease and concentrate only at the vicinity of the bend. The gas holdup distributions shown in Fig. 6 are presented in greater detail in separate plots in Figs. 7 9 for each specific flow regime. In plug flow (Fig. 7), distinct features are associated with each flow direction. In upward flow, the gas holdup enters the measuring section with a fairly constant value, but starts to decrease some 25 diameters upstream of the bend as a result of the flow deceleration brought about by the change in flow direction. In the bend, a change in gas holdup is observed due to the expansion and acceleration of the Taylor bubbles. This is caused by the sudden change in static pressure, mainly as a result of the change in static head experienced by the bubble while it flows upward in the bend (as depicted in Fig. 1). For this reason, the gas holdup of developed flow far downstream of the bend is higher than at the upstream region. Liquid flow reversal (countercurrent flow) was systematically observed in the bend for the upward condition (de Oliveira and Barbosa, 214). This contributes to the high values of gas holdup at the outlet (top) of the bend, as well as to the low gas holdup values at the inlet (bottom). The intermittent liquid accumulation in the bottom region of the bend reduces the timeaveraged gas holdup that is measured by the inlet gas holdup sensor. As a result of the changes in flow structure due to the bend, significant variations in holdup (in a qualitative sense) were observed within 4 diameters upstream and 7 diameters downstream of the bend, approximately. Video sequences of the present experiments can be accessed at airwater-returnbends. Although points of maximum gas holdup in upward and downward flow occur at the outlet of the bend (Fig. 7), they are caused by different physical phenomena. In the case of downward plug in out α [-] in out Upward Downward * Fig. 8. Gas holdup distribution for upward and downward slug flow ( j l =.2 m/s, j g = 4 m/s). α [-] in out Upward Downward * Fig. 9. Gas holdup distribution for upward and downward annular flow ( j l =.2 m/ s, j g = 2 m/s)..7.6 α [-] Upward Downward * z ref [m] Fig. 7. Gas holdup distribution for upward and downward plug flow ( j l =.2 m/s, j g =.4 m/s). Fig. 1. Phase distribution in upward plug flow ( j l =.2 m/s, j g =.4 m/s). The sequence illustrates the nose and body of a rising bubble, with a clear occurrence of liquid flow reversal in the bend in the third frame. The time interval between frames is 152 ms.

7 38 P.M. de Oliveira et al. / International Journal of Multiphase Flow 67 (214) flow, the liquid phase is accelerated towards the bottom part of the bend because of gravity. This effect was also observed in downward slug and annular flows, though in a smaller scale than in plug flow. The latter flow regime is depicted in Fig. 11, which shows the descent of a Taylor bubble in the return bend at two instants: when the bubble nose is at the inlet (left), and when the nose is at the outlet (right). As a result of the increase in the phase velocities at the outlet, a decrease in the time-average liquid holdup is observed at the bottom of the bend. In contrast with upward plug flow, a slight increase in gas density is expected due to the larger static head. This contributes to the gas holdup being smaller in the bottom part. Fig. 12 shows three distinct instants of the descent of the tail of a Taylor bubble in the bend. Since buoyancy acts against the net flow, the bubble is decelerated in the bend. This, combined with interfacial friction and centrifugal effects, gives rise to a breakup of the gas liquid interface and large disturbances that propagate downstream along the bottom tube. The developing length downstream of the bend seems to extend beyond 1 diameters for the gas holdup in this case. In the slug and annular flow regimes (Figs. 8 and 9), a much smaller axial variation was observed. The values of gas holdup farthest downstream remained within approximately 7% and 1% of the initial upstream value for the conditions of slug and annular flow, respectively. As the superficial gas velocity increases and the flow regime changes from slug to annular, the gas holdup distributions for both flow directions become more similar to each other due to a predominance of the inertial effects. In Fig. 9 (annular flow), the differences in gas holdup between the upward and downward flows are confined to approximately 3 diameters to the inlet and from the outlet of the bend. However, it should be noted that the amplitude of the gas holdup perturbations in annular flow are of the same order of magnitude as the experimental uncertainty see electronic annex. In slug flow, the differences in holdup between the two flow directions persist well beyond the 3-diameter limit. In qualitative terms, however, the gas holdup distributions for slug flow are similar to those for plug flow, indicating that centrifugal and buoyancy forces still influence quite significantly the flow behavior of the phases in the bend Pressure gradient Friction and flow acceleration In addition to the changes in gas holdup, the flow development upstream and downstream of the bend was also investigated through an assessment of the friction and acceleration components of the axial pressure gradient. Fig. 12. Phase distribution in downward plug flow ( j l =.2 m/s, j g =.4 m/s). The sequence shows the tail of a descending bubble, followed by roughening and breakup of the gas liquid interface. The time interval between frames is 91 ms in out p f -8 p a p f + p a (a) in out Fig. 11. Phase distribution in downward plug flow ( j l =.2 m/s, j g =.4 m/s). The sequence shows the nose and body of a descending bubble. The time interval between frames is 532 ms (b) Fig. 13. Frictional and accelerational pressure gradients in (a) downward and (b) upward plug flow ( j l =.2 m/s, j g =.4 m/s).

8 P.M. de Oliveira et al. / International Journal of Multiphase Flow 67 (214) The components of the pressure gradient due to friction and acceleration are shown in Figs for plug, slug and annular flow. The bottom horizontal axis shows the distance in relation to the reference position at the inlet of the pressure drop measuring section (z ref ). Again, the upper horizontal axis represents the distance in diameters to the inlet and from the outlet of the bend. The change in pressure due to the static head was subtracted from the total pressure change in the bend (Eq. (3)) to permit a better visualization of the gradients of the other terms. Fig. 13 shows the pressure gradients for plug flow in the (a) downward and (b) upward directions. In downward flow, the frictional pressure gradient is approximately constant up to 1 diameters upstream of the bend, when gravity accelerates the liquid down the bend, thus increasing the average wall shear stress. The negative peak in the frictional pressure gradient is due to friction and recirculation losses in the bend and, as the flow enters the bottom tube, is restablished to virtually the same values upstream of the bend after 3 5 diameters or so. The largest variation of the accelerational pressure gradient occurs immediately downstream in out in out (a) in out p f z ref [m] 2 (a) in out p a p + p f a (b) Fig. 15. Frictional and accelerational pressure gradients in (a) downward and (b) upward annular flow ( j l =.2 m/s, j g = 2 m/s) p f p a p + p f a z ref [m] (b) Fig. 14. Frictional and accelerational pressure gradients in (a) downward and (b) upward slug flow ( j l =.2 m/s, j g = 4 m/s). of the bend, where it sharply changes from negative to positive. Far from the bend, where the frictional component is almost constant, the influence of inertia is almost negligible, as expected. The decreasing trend of the accelerational pressure gradient upstream of the bend is easily justified by the acceleration of the liquid in the bend, and the consequent thinning of the liquid film in the long Taylor bubbles upstream of the bend (see Fig. 11). The change in sign of the inertia component of the pressure gradient coincides with the axial reduction of the gas holdup after the bend (see Fig. 7), which occurs as a result of the deceleration of the liquid film and breakup of the tail of the Taylor bubbles into smaller parts. Overall, the flow pressure drop seems to be recovered at about 6 diameters downstream, with the fading of the reversible pressure changes. In upward plug flow, the frictional pressure gradient is strongly influenced by the liquid flow reversals in the bend, as can be seen in Fig. 13b. Immediately upstream of the bend, the frictional and accelerational components reach their maximum values, but with opposite signs. The (positive) peak in rp a results from the

9 4 P.M. de Oliveira et al. / International Journal of Multiphase Flow 67 (214) (periodic) liquid accumulation at the bottom of the bend, which increases the liquid holdup (see Fig. 7) and creates recirculation in the liquid, thus giving rise to friction losses. The time-average frictional pressure gradient at the outlet of the bend is positive due to periodic reversals of the liquid, which falls back in the bend, creating a flow condition that resembles the behavior of churn flow in a vertical pipe (Hewitt et al., 1985), with a chaotic mixing of the liquid in the bend, as illustrated in Fig. 1. At this flow condition, flow recovery takes place at around 6 diameters downstream, which is almost equivalent to the flow length affected by the bend upstream of it. The behavior of downward slug flow (Fig. 14a) is qualitatively very similar to downward plug flow for both the frictional and accelerational components. Nevertheless, flow recovery seems to occur at a shorter length downstream (approximately 3 diameters), and no influence is perceived 3 diameters upstream of the bend. Virtually the same upstream and downstream limits apply to upward slug flow (Fig. 14b). In this case, although some liquid flow reversal was observed in the bend, it was not sufficient to change the sign of the frictional component of the pressure gradient due to the larger inertia of the gas (in comparison to the plug flow condition). Fig. 16 shows the phase distribution for slug flow in the upward (left frame) and downward (right frame) directions, where each frame shows the liquid slug entering the curved section. In the left frame, significant mixing of the flow is observed due to the descending liquid, while in the right frame the liquid phase is concentrated mostly in the outer part of the bend. As far as the annular regime is concerned (Fig. 15), both pressure gradient distributions indicate that this flow regime is the least affected by the bend, due to the predominance of inertia effects. Still, some slight differences occur. In both orientations, the maximum absolute value of the accelerational pressure gradient takes place at the vicinity of the bottom turn of the bend. In the downward flow, the liquid concentrated in the bottom part of the disturbance waves traveling in the upper horizontal tube is detached from the bottom film towards the outer part of the curve as a result of centrifugal forces. This behavior is depicted in Fig. 17; in the left frame, the dark region located between the middle and the top of the bend is a disturbance wave being ejected towards the outer wall. This wave appears in the right frame (after 3 ms) as a dark region at the bottom portion of the bend. In the upward annular flow, as seen in Fig. 18, the film is already thicker at the outer part of the curve when the flow enters the bend. At the upper part of the bend, the liquid is gradually redistributed from the upper region of the tube to the bottom due to gravity. In both flow directions, the resulting effect is a disintegration of the disturbance waves in annular flow, as demonstrated in the Harwell high-speed film of two-phase flow in return bends (George, 1971). In other Fig. 17. Phase distribution for downward annular flow in the bend ( j l =.2 m/s, j g = 2 m/s). Time interval between frames is 3 ms. Fig. 18. Phase distribution for upward annular flow in the bend ( j l =.2 m/s, j g = 2 m/s). words,...the bend destroyed the disturbance waves in annular flow. These waves were essential to keeping the top wet so (immediately after the bend) dryout could occur in heated systems... (Hewitt, 213). 4. Conclusions In order to quantify the influence of a vertical 18 return bend on the development of air water flows in horizontal tubes, gas holdup and pressure gradient data were obtained as a function of distance in a 26.4-mm ID test section. The experiments were carried out for the plug, slug and annular flow regimes, covering both upward and downward directions. Frictional and accelerational components of the total pressure gradient were evaluated for j l =.2 m/s and j g ranging from.2 to 2 m/s. The results showed that the plug flow regime was more affected by gravitational effects than the slug and annular flow regimes, for which inertial and frictional effects were dominant. According to the gas holdup measurements, the influence of the bend was observed at around 7 diameters upstream of the bend, and was seen to extend beyond the measuring limits (more than 1 diameters) in downward plug flow. Nevertheless, significant changes in pressure gradient were confined to 4 diameters upstream and 6 diameters downstream of the bend. Acknowledgements Fig. 16. Phase distribution in upward (left) and downward (right) slug flow in the bend ( j l =.2 m/s, j g = 4 m/s). The authors would like to thank Prof. M. da Silva, N. Libert, and L. Lipinski (UTFPR) for providing the hardware of the gas holdup

10 P.M. de Oliveira et al. / International Journal of Multiphase Flow 67 (214) measurement system, P. Cardoso (UFSC) for his craftsmanship in constructing the gas holdup probes and Helder Martinovsky (Unisul) for the 16-mm film conversion. Financial support from Petrobras and CNPq through Grant No /28-8 (National Institute of Science and Technology in Cooling and Thermophysics) is gratefully acknowledged. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at References da Silva, M.J., Schleicher, E., Hampel, U., 27. Capacitance wire-mesh sensor for fast measurement of phase fraction distributions. Meas. Sci. Technol. 18, da Silva Lima, R.J., Thome, J.R., 21. Two-phase pressure drops in adiabatic horizontal circular smooth U-bends and contiguous straight pipes (RP-1444). HVAC&R Res. 16, da Silva Lima, R.J., Thome, J.R., 212. Two-phase flow patterns in U-bends and their contiguous straight tubes for different orientations, tube and bend diameters. Int. J. Refrig. 35, De Kerpel, K., De Groof, C., Dreesen, M., De Keulenaer, T., De Paepe, M., 213. Capacitive sensor measurements on two-phase flow in smooth return bends. In: Proceedings of the 8th Experimental Heat Transfer, Fluid Mechanics and Thermodynamics June, Lisbon. de Oliveira, P.M., Barbosa Jr., J.R., 214. Pressure drop and gas holdup in air water flow in 18 return bends. Int. J. Multiphase Flow 61, George, K.K., Two-Phase Flow in 18 Return Bends High Speed Cine Film, Report AERE-M 2459, UK AEA. Hewitt, G.F., 213. Private Communication. Hewitt, G.F., Martin, C.J., Wilkes, N.S., Experimental and modelling studies of annular flow in the region between flow reversal and the pressure drop minimum. Phys. Chem. Hydrodyn. 6, Hoang, K., Davis, M.R., Flow structure and pressure loss for two phase flow in return bends. J. Fluids Eng. 16, 3. Libert, N., 213. Sistema de Medição Capacitivo para Determinação da Fração de Vazio em Escoamentos Bifásicos. Master Thesis, Universidade Tecnologica Federal do Paraná. Libert, N., Lipinski, L., da Silva, M., 211. Capacitive probe for gas liquid flow characterization. In: Proceedings of XVIII IMEKO TC4 Symposium and IX Semetro. Natal. Padilla, M., Revellin, R., Wallet, J., Bonjour, J., 213. Flow regime visualization and pressure drops of HFO-1234yf, R-134a and R-41A during downward twophase flow in vertical return bends. Int. J. Heat Fluid Flow 4, Smith, S.L., Void fractions in two-phase flow: a correlation based upon an equal velocity head model. Proc. Inst. Mech. Eng , Steiner, D., Verein Deutscher Ingenieure VDI-Wärmeatlas, chapter HBB. VDI- Gesellshaft Verfahrenstechnik und Chemieingenieurwesen (GVC), Düsseldorf. Traviss, D., Rohsenow, W., The influence of Return Bends on the Downstream Pressure Drop and Condensation Heat Transfer in Tubes. Tech. Rep. MIT Heat Transfer Laboratory, Cambridge, Massachusetts. Usui, K., Aoki, S., Inoue, A., 198. flow behavior and pressure drop of two-phase flow through C-shaped bend in vertical plane, (I) upward flow. J. Nucl. Sci. Technol. 17, Usui, K., Aoki, S., Inoue, A., Flow behavior and pressure drop of two-phase flow through C-shaped bend in vertical plane, (II) downward flow. J. Nucl. Sci. Technol. 18, Wang, C.-C., Chen, I.Y., Yang, Y.-W., Chang, Y.-J., 23. Two-phase flow pattern in small diameter tubes with the presence of horizontal return bend. Int. J. Heat Mass Transf. 46, Wang, C.-C., Chen, I.Y., Yang, Y.-W., Hu, R., 24. Influence of horizontal return bend on the two-phase flow pattern in small diameter tubes. Exp. Therm. Fluid Sci. 28, Wang, C., Chen, I., Huang, P., 25. Two-phase slug flow across small diameter tubes with the presence of vertical return bend. Int. J. Heat Mass Transf. 48, Wang, C., Chen, I., Lin, Y., Chang, Y., 28. A visual observation of the air water twophase flow in small diameter tubes subject to the influence of vertical return bends. Chem. Eng. Res. Des. 86,