Measurement of dynamic void fractions in stratified

Experimental Thermal and Fluid Science 29 (2005) 383–392 www.elsevier.com/locate/etfs

Measurement of dynamic void fractions in stratified types of flow Leszek Wojtan *, Thierry Ursenbacher, John R. Thome Laboratory of Heat and Mass Transfer (LTCM), Swiss Federal Institute of Technology Lausanne (EPFL), CH-1015 Lausanne, Switzerland Received 14 September 2003; received in revised form 27 April 2004; accepted 21 May 2004

Abstract A new non-intrusive computerized image analysis and optical observation method for accurately measuring dynamic void fraction has recently been developed by Thome and coworkers [Dynamic void fractions in stratified types of flow, part I: new optical measurement technique, Int. J. Multiphase Flow 31 (2004); Dynamic void fractions in stratified types of flow part II: measurements for R-22 and R-410a, Int. J. Multiphase Flow 31 (2004)]. This technique is applied to circular horizontal sight glass tubes using a monochromatic laser sheet to illuminate the two-phase flow coupled with image processing to measure cross-sectional void fractions and dry angles in stratified types of flow. The refraction effects on the cross-sectional images are overcome by reconstructing the video images by computer. From these, the shape at the vapor–liquid interface is detected and the void fraction is accurately determined (to an estimated accuracy of about ±0.01) over a wide range of void fraction values (from 0.05 to 0.95). In addition, the dry angle around the upper perimeter of the tube is also obtained. The system has been coupled to a flow boiling test facility to obtain dynamic and time-averaged void fractions in a horizontal tube for two refrigerants: R-22 and R-410A. About 227,000 images were analysed so far in this study for a 13.6 mm tube to provide the same number of dynamic void fraction measurements and 238 timeaveraged void fraction values. A summary of the technique is described here with additional analysis of the previous results and new results (based on processing 83,000 new images) are presented here for void fractions in an 8.0 mm sightglass tube with R-22 at mass velocities of 100 and 150 kg/(m2 s). The new method also has the potential to measure interfacial wave contours and other pertinent geometrical characteristics of stratified two-phase flows. 2004 Elsevier Inc. All rights reserved.

1. Introduction Completely empirical flow boiling heat transfer design methods have recently been succeeded by flow pattern/ flow structure based, boiling heat transfer models, which has yielded significant improvements in the accuracy of two-phase flow heat transfer predictions. Prior methods did not account for the effects of two-phase flow regimes and were especially inaccurate at high vapor qualities and for stratified types of flows. In 1998, Kattan et al. [3–5] proposed a new flow pattern map for adiabatic and evaporating flows in small diameter horizontal tubes based on a new large database of flow patterns for five refrigerants and presented a new flow boiling heat trans*

Corresponding author. Fax: +41 21 693 5960. E-mail address: john.thome@epfl.ch (L. Wojtan).

0894-1777/$ - see front matter 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.expthermflusci.2004.05.017

fer model that assumed simplified two-phase flow structures based on this flow pattern map. This method successfully predicted local heat transfer coefficients in the stratified, stratified-wavy, intermittent, annular and annular flow with partial dryout regimes. More recent yet, an analogous in-tube condensation, flow pattern based heat transfer model based on a condensation version of their same flow pattern map has been proposed by Thome and coworkers [6,7], including the effect of interfacial roughness, that predicts 85% of a 2700+ database obtained from nine independent laboratories to within ±20% for a list of 15 fluids and a very wide range of operating conditions. Hence, two-phase heat transfer prediction models, which account for flow regime effects, are considered state of the art and have the potential to significantly increase accuracy and reliability. Developing these methods from one flow pattern map, this leads

384

L. Wojtan et al. / Experimental Thermal and Fluid Science 29 (2005) 383–392

Nomenclature D R acb adry anb

internal round tube diameter, m internal round tube radius, m convective boiling heat transfer coefficient, W/(m2 K) vapor heat transfer coefficient, W/(m2 K) nucleate boiling heat transfer coefficient, W/ (m2 K)

to a unified approach to two-phase heat transfer. On the other hand, further improvements to these models require precise information on flow pattern transitions, two-phase flow structures, mean phase velocities, void fractions, dry angles in stratified types of flows, liquid film thickness and interfacial roughness. The most important parameter influencing two-phase flow characteristics is the cross-sectional void fraction e. For stratified flows, the dry angle hdry is equally important because it delineates the wet and dry perimeters for heat transfer and pressure drop. According to the simplified flow structures assumed to represent annular, stratified-wavy and stratified flows as annular films as shown in Fig. 1, the Kattan–Thome– Favrat flow boiling heat transfer model for evaporation in horizontal tubes is obtained as local perimeter average of the heat transfer coefficients around the tube perimeter as: atp ¼

hdry adry þ ð2p hdry Þða3nb þ a3cb Þ 2p

1=3

ð1Þ

Fig. 1. Flow structures for annular, stratified-wavy and fully stratified flows (left to right in the bottom three diagrams) and for fully stratified flow and its film flow equivalent (top two diagrams).

atp d e hdry qL qV

two-phase flow heat transfer coefficient, W/ (m2 K) liquid film thickness, m cross-sectional void fraction dry angle, rad density of the saturated liquid phase, kg/m3 density of the saturated vapor phase, kg/m3

where atp is the perimeter averaged two-phase heat transfer coefficient, anb, is the nucleate boiling heat transfer coefficient, acb is the convective boiling heat transfer coefficient and adry is the vapor heat transfer coefficient. As can be seen in Eq. (1), for stratified types of flow the dry angle is an important parameter affecting heat transfer in flow boiling and has a similar importance to condensation and two-phase pressure drops. Similarly, the void fraction is the key parameter in predicting twophase flow pattern transitions and the values of acb and hdry. In the present study, a recently completed optical image analysis technique [1,2] utilizing a laser sheet and a video camera to determine the two-phase interface inside round horizontal glass tubes is first presented, then additional analysis of the previous results for a 13.6 mm sightglass tube, and finally some new void fraction data for a small 8.0 mm sightglass tube. The technique is applicable only to stratified types of two-phase flows, in particular: fully stratified flow, stratified-wavy flow and annular flow with partial dryout around the top half of the tube. Slug flows have also been successfully measured as long as the liquid slug leaves behind either no liquid film on the upper perimeter of the tube or only an extremely thin film that does not disturb the optical view of the stratified liquid below. Since the void fraction is determined from the locus of the stratified liquid–vapor interface, it is only determinable if no bubbles are in the liquid phase. Hence, test results are only reported for conditions in which no bubbles are visible. At the last 2nd Japanese–European Two-Phase Flow Group Meeting, Tsukuba, Japan in 2000, Thome et al. [8] presented a description of our first stage of development of this new dynamic void fraction/dry angle measurement technique. That was essentially a proof-of-concept exercise. The technique has since been perfected by Thome and coworkers [1,2] as noted above. Our specific objective is to measure void fractions for these stratified types of flows in tubular sight glasses located at the end of heat transfer test sections. This allows simultaneous observation of flow patterns and measurement of dry angles, void fractions, heat transfer coefficients and two-phase pressure drops in a single test


facility for experimental studies on in-tube boiling and condensation. In the future, we also plan to use longitudinal views of the flow from the side of the tube to obtain valuable interfacial roughness measurements, an advancement that requires a higher definition video camera. Other methods are available for measuring crosssectional void fractions in two-phase flows: resistance sensors [9], capacitance or conductance sensors [10], refractive fiber optic probes [11], radiation attenuation techniques [12], ultrasonic tomography [13], etc. With intrusive methods, all types of sensors or probes disturb to some extent the flow field and induce imperfect identification of the interface between the gas and liquid. Non-intrusive methods, such as radiation techniques and conductance measurements, yield mean or cordal void fraction values of the cross-section. Using these methods, one cannot determine either the dry angle nor the shape of the interface between the gas and liquid. More sophisticated non-intrusive methods, such as real time neutron radiography (RTNR) and X-ray computed tomography (X-CT), allow for the measurement of the distribution of void fraction. Notably, Harvel et al. [14] are able to reconstruct two-dimensional cross-sectional images and determine dynamic void fractions. The present laser image processing technique to be described here for stratified types of flows appears to be the most accurate currently available.

2. Optical image processing analysis technique Fig. 2 depicts the optical setup for a borosilicate glass tube installed at the end of the heat transfer

Fig. 2. Schematic of the optical setup.

385

measurement test section. The internal diameter of the glass tube is 13.6 mm whose wall thickness is 1.2 mm (±0.02 mm); the second tubular sightglass tube tested more recently is 8.0 mm in internal diameter with a 1.0 mm wall. A cross-section of the glass tube and the fluid within are illuminated by a laser sheet (Spectra Physics laser Millennia II) and the images are recorded by a digital camera (Panasonic CCD camera type GP-LM7/TA) above the glass tube with a vision angle of 40 with respect to the axis of the tube. The laser sheet is about 0.5 mm thick. The wavelength emitted by the laser is 532 nm and interlaced images at 25 Hz are recorded on a Pentium 3 PC through a Targa Video card. These images are then unlaced and hence provide 50 images/s. The image of the two-phase flow inside the glass tube in Fig. 2 is distorted by the refraction of the light passing through the internal and external surfaces of the glass. Thus, this technique requires reconstruction of the field of the vision to obtain a non-deformed, orthogonal view of the illuminated cross-section. This reconstruction is done by placing a regular grid inside the glass tube in the same plane as the laser sheet during preliminary tests without any liquid in the tube. The process consists of extracting an inverse transformation function, which is applied to the distorted grid image in order to recreate the original image of a real grid. This function is then applied to the video images of the twophase refrigerant flowing in the tube. The transformation, however, is only valid for the cross-sectional fraction occupied by a gas under the condition that the gas phase is in direct contact with the upper wall of the tube; a liquid film on the top of the tube, otherwise, would create yet another refraction of the light. Furthermore, the transformation is only valid in the gas phase above the vapor–liquid interface and not below into the liquid phase (where the light is again refracted). For this reason, the new technique is only applicable to stratified types of flow with vapor in the upper zone of the tube and without entrained bubbles in the liquid phase (undetectable by the method). The locus of the interface is detected using image processing, giving the ratio of the area occupied by the vapor phase with the total cross-sectional area, i.e. the instantaneous void fraction as a function of time, and the dry angle. Our inhouse image processing program is coded within Lab View (version 6i). A high definition grid of known dimensions printed on glossy paper is fixed inside the glass tube in a calibration set-up (depicted in Fig. 3) identical to that on the flow boiling test facility. Once the optical alignment of the laser and the camera setup are made, it remains unchanged whether the upper support is mounted on the calibration set-up or on the flow boiling test facility. The grid is used to model the optical distortion and to calibrate the interface detection procedure between the

386


Fig. 5. Transformed grid image. Fig. 3. Image calibration and video acquisition system setups.

gas and liquid. Such an optically distorted grid inside the sight glass tube is depicted in Fig. 4. The distorted grid image is converted into an 8-bits encoded image with 256 levels of grey. Due to the small distance between the camera and the grid (about 3 cm) and the vision angle of 40, the depth of focus is short, making it impossible to obtain a well-focused image of the total surface of the grid. The focal plane is thus adjusted on the horizontal grid axis such that the lines above and below the axis do not become too blurred. The zones outside of the grid are eliminated by decreasing the image size from 720 · 576 pixels to 430 · 370 pixels and by adding an elliptical mask to focus only on the internal surface of the glass tube (Fig. 4). The image is also inverted (in terms of gray levels) allowing an easier distinction of the grid border. Optically, two types of distortion can be distinguished on the deformed grid image: the bottom part of the grid (under the axis) is vertically compressed (diminution of the space between the horizontal lines) and the upper part is horizontally elongated (increasing space between the vertical lines). The result of the image transformation is illustrated in Fig. 5. Looking in the plane of the drawing, the horizontal and vertical lines in the transformed grid image are nearly perfectly straight and parallel to one another, even near the perimeter. Thus, it is now possible to

map, pixel by pixel, the location on the distorted image back to the real plane and hence apply this to two-phase flow videos frame-by-frame to reconstruct the video image for further processing for detection of the twophase interface. A typical evolution of a luminance signal is represented in Fig. 6, with information extracted from the median pixel column of a transformed image corresponding to a void fraction of 55% and a laser intensity level of 61%. The transition from the gas zone in which the luminance L is on the order of 50 bits to the liquid zone where L @ 200 bits occurs over 10–15 pixels. It is necessary to fix a luminance threshold above which a gas phase is considered present, and inversely, below which there is a liquid phase. In terms of image processing, selecting only one pixel with a luminance above the threshold is sufficient to consider all the contiguous pixels also with a luminance above the threshold as belonging to the liquid cross-sectional area (this corresponds to the well-known function called magic wand). Thus the upper border of this zone represents the liquid–vapor interface. Selecting all the pixels above this limit but inside the mask, we obtain the representative surface of the vapor phase. Finally, dividing the number of pixels, which represents this vapor phase with the total number of pixels that are inside the mask, one finds the void fraction. Similarly, the dry perimeter around the upper perimeter of the tube can be determined.

Fig. 4. Original (left) and limited (right) grid image (the white transversal line corresponds to the 230th line, counted from the top on the limited image).


387

Fig. 6. Transformed image (luminance plane) of the illuminated cross-section (left), luminance intensity over the median vertical line (center) and extracted vapor zone corresponding to a fixed threshold value (right).

Based on calibrations of the technique versus static two-phase interfaces whose void fractions were accurately determined from gravimetric balance measurements, the optimum luminance thresholds were determined based on laser light intensity. The void fraction measurement error of this new optical technique is estimated to be on the order of ±0.01 in void fraction over the range of void fractions measurable by this technique for stratified types of flow (0.05 < e < 0.95). This represents a very accurate method applicable of nearly the entire range of void fractions.

3. Dynamic void fraction measurements The dynamic void fraction measurements for refrigerant R-410A in a stratified-wavy/intermittent flow condition are plotted versus time in Fig. 7 at a mass velocity of G = 70 kg/(m2 s), a vapor quality of x = 0.2 and a saturation temperature of 5 C. The characteristic cyclic variation observable corresponds to waves passing through the cross-section and the time in-between corresponds to their period. The void fractions equal to 0.0 represent the passage of liquid slugs that fill the entire channel. Fig. 8 shows cross-sectional images within the

Fig. 7. Void fraction evolution for stratified-wavy/intermittent flow.

time interval of 9.50–11.16 s, which correspond to passage of a wave with respective void fractions of 0.537, 0.685, 0.794 and 0.479 in Fig. 7. There are two groups of images corresponding to two different steps of the image processing. The first group (top) is obtained after optical transformation and interfacial detection (black curves). As can be seen, the position of interface is precise and parameters like the dry angle and the liquid height can be very accurately determined. The second group (bottom) illustrates the final black and white processed images used for void fraction calculation. The cross-sectional area below the gas/liquid interface is filled with black color and represents the liquid phase. Inversely, the white color corresponds to the gas phase. The cross-sectional void fraction is thus calculated from the ratio of the white pixels to the total number of pixels of the channel. Fig. 9 shows representative graphs of the results obtained for R-410A and R-22 at one of the four mass velocities tested compared to four void fraction models. The tubular sightglass is 13.6 mm in internal diameter and fits tightly to the end of a 13.84 mm diameter evaporator tube. The upper curve depicts the homogeneous void fraction (shown for reference purposes only as the upper feasible limit). The second curve is for the Rouhani and Axelsson [15] void fraction model modified by Steiner [16] for application to horizontal flows: " x x 1x e¼ ð1 þ 0:12ð1 xÞÞ þ qV qV qL # 0:25 1 1:18ð1 xÞ½grðqL qV Þ þ ð2Þ Gq0:5 L This is a type of a drift flux model. Steiner did not provide a comparison of Eq. (2) to void fraction data but only noted that he found it to work for R-12 and R-22. Kattan et al. [5] used this expression in their twophase flow heat transfer model described earlier and more recently also in the new condensation heat transfer model and flow pattern map of Thome and coworkers [6,7]. This expression provides a simple method for

388


Fig. 8. Movement of the liquid interface during one wave cycle (from 9.50 to 11.16 s): transformed images with detected interface (upper part) and final images (bottom part).

Fig. 9. Void fractions of R-410A and R-22 for a 13.6 mm sightglass tube compared to four models.

calculating void fractions including the effects of mass velocity and surface tension. Two additional void fraction models have been added to the comparison: Zivi [17] and Taitel and Dukler [18]. Zivi derived his model based on minimizing the kinetic energy of the twophases assuming no liquid entrainment in annular flows while Taitel and Dukler developed a method to predict the liquid height in stratified flows for their well-known flow pattern map that can also be used to obtain the void fraction. As can be seen, this void fraction equation is quite successful in representing the experimental time-averaged void fraction values, with similar success displayed at six other mass velocities. Each time-averaged void fraction value is obtained from about 900 dynamic values; thus the data in Fig. 9 represents about 50,000 processed images! Fig. 10 presents new results obtained for R-22 at 5 C (as the others) at two mass velocities that are compared

to the same four void fraction models. The tubular sightglass is 8.0 mm in internal diameter and fits tightly to the end of an 8.0 mm diameter evaporator tube. A new optimum grid and the calibration process for the image processing versus static void fraction measurements using a gravimetric balance were repeated here for this 8.0 mm tube as was described for the 13.6 mm tube in [1], Here, flow visualization was much more difficult than for the larger tube, in part because the number of pixels per image are less for the smaller tube. Furthermore, getting images suitable for image processing was more difficult due the higher curvature of the smaller tube. In fact, while video images were obtained for both R-22 and R-410A at four mass velocities during the heat transfer experiments for the 8.0 mm tube similar to those for the larger tube, only part of two series for R-22 proved to be suitable for image processing. As opposed to the larger tube, in the 8.0 mm tube liquid slugs and high amplitude waves tended to leave behind a


389

Fig. 10. Void fractions of R-22 for an 8.0 mm sightglass tube compared to four models.

liquid film on the upper perimeter of the tube. This film is thicker compared to those occasionally appearing in the 13.6 mm sightglass tube and causes a problem for the stratified liquid interface detection technique. For example, Fig. 11 shows a representative image of such a situation for the 8.0 mm tube for R-22 at a mass velocity of 150 kg/(m2 s) and vapor quality of 0.676 with successive steps of image processing. As can be seen in the left image, the stratified liquid interface in the lower part of the tube is visible but is badly blurred by the upper film of liquid (bright white arc around upper perimeter), and when transformed in the other two images, the stratified liquid interface cannot be detected (or incorrect data are obtained). Hence, videos sequences had to be visually processed first (the eye is better than any image processing system for detecting irregularities) to determine where the limit of the method was reached. The vertical lines on the two graphs in Fig. 10 indicate this threshold for these two test series. Hence, all the data to the right of the lines are not valid and are only shown to illustrate the problem. The presence of a liquid film on the upper perimeter of the tubular sightglass creates an additional refraction in viewing of the interface by the video camera. If this film is extremely thin and its interface is smooth, then the measurement technique

is not noticeably affected. However, if the film is not so thin and not smooth, then the location of the stratified liquid cannot be determined and the image is also blurred. Consequently, this is a limitation of the range of application of this technique. Tables 1 and 2 depict the details of the new comprehensive statistical comparison of the 13.6 and 8.0 mm sightglass void fraction data to four leading models. For the 13.6 mm data, the values of the average, mean and standard deviation of all the data are 8.5%, 12.0% and 20.2% for the Taitel–Dukler model while 1.5%, 7.7% and 14.1% for the Rouhani–Axelsson model (most of the larger errors come from the void fraction measurements at vapor qualities less than 0.05, where the effects of the energy balance on determining the experimental vapor quality are magnified). This is a very good result and qualifies the Rouhani–Axelsson method as a very accurate one for building flow transition, heat transfer and two-phase pressure drop prediction methods. Similar comments can be made about the statistical comparison for the 8.0 mm tube, i.e. 6.1%, 6.9% and 9.2% for the Rouhani–Axelsson model, for which only the values to the left of the vertical lines in Fig. 10 were used. Hence for the 8.0 mm sightglass, here the Rouhani–Axelsson model as modified by Steiner works

Fig. 11. Images for R-22 in the 8.0 mm sightglass tube showing a sequence of images: left = original image, middle = transformed image, right = blackand-white image processed for void fraction (no white means no voids here even though the stratified liquid is visible in the original image).

390


Table 1 Statistical comparison of the 13.6 mm sightglass data to the four void fraction models Test conditions

Model

Average deviation (%)

Mean deviation (%)

Standard deviation (%)

R-22, 70 kg/(m s)

Homogeneous Zivi Taitel–Dukler Rouhani

22.0 1.2 6.8 1.2

16.7 7.5 10.4 6.6

28.2 13.0 18.2 11.7

R-22, 100 kg/(m2 s)


15.8 5.8 0.1 4.1

10.5 9.6 3.5 4.8

13.6 12.8 5.6 7.7

R-22, 150 kg/(m2 s)


14.7 6.7 0.4 1.6

13.0 8.9 3.6 3.3

17.8 11.9 7.3 6.7

R-22, 200 kg/(m2 s)


18.0 3.8 3.4 2.9

17.4 7.7 7.6 7.6

25.7 15.1 13.0 15.1

R-410A, 70 kg/(m2 s)


47.2 8.7 30.6 7.0

43.4 16.3 37.5 16.6

77.6 32.1 61.8 33.8

R-410A, 150 kg/(m2 s)


20.4 6.8 6.5 0.1

15.6 13.0 7.7 7.4

20.7 18.4 12.6 14.3

R-410A, 200 kg/(m2 s)


14.9 8.6 3.7 1.1

13.9 15.9 6.4 9.8

22.5 23.1 14.7 18.8

R-410A, 300 kg/(m2 s)


31.5 9.0 17.3 9.6

16.3 12.5 19.3 5.5

21.9 15.0 28.5 8.8

Overall results


23.1 3.9 8.5 1.5

18.3 11.4 12.0 7.7

28.5 17.7 20.2 14.6

2

well for this predominantly low vapor quality range of data, even though they are the most difficult ones to measure and predict.

4. Dry angle measurements Fig. 12 depicts the new dry angle measurements obtained with R-22 for the 8.0 mm tube compared to those for the 13.6 mm tube presented in [1,2]. The dry angles shown are time-averaged values. At low vapor qualities there tends to be a large cyclic variation in the dynamic dry angle with time since there are large amplitude

waves. The accuracy of the measured values is about ±10, except at high vapor qualities where the two ends of the thin liquid arc remaining in the tube cannot be detected accurately. The maximum feasible dry angle can be calculated by assuming a completely stratified flow (horizontal and flat interface) and applying the void fraction equation shown in Eq. (2) to the area occupied by the vapor and thus determine hdry. As can be seen, most of the measured values are equal to or less than this maximum value, and in fact as a first estimate are quite well represented by this simple prediction. The measured values for x > 0.75 tend to fall off in value at high vapor qualities because of limitations in the meas-


391

Table 2 Statistical comparison of the 8.0 mm sightglass data to the four void fraction models Test conditions

Model

Average deviation (%)

R-22, 100 kg/(m s)


19.3 4.2 4.8 2.9

3.6 7.2 4.4 4.9

4.5 10.1 5.7 6.7

R-22, 150 kg/(m2 s)


25.7 6.0 8.9 9.2

3.9 15.0 9.4 8.9

5.6 18.5 11.9 11.8

Overall results


22.5 5.1 6.8 6.1

3.8 11.1 6.9 6.9

5.0 14.3 8.8 9.2

2

Mean deviation (%)

Standard deviation (%)

Fig. 12. Measured dry angle compared to stratified angle as a function of vapor quality for R-22 at 100 and 150 kg/(m2 s).

urement technique as noted above, and these values are thus only approximate indications of their real values (±30).

5. Conclusions A recently developed optical measurement technique to dynamically measure the location of the vapor–liquid interface in two-phase stratified flows in horizontal tubes

is described and some new results for an 8.0 mm sightglass tube are shown in comparison to the previous results for 13.6 mm. The new method is non-intrusive and allows measurements of the dry angle and void fraction through a glass tube in a cross-sectional view perpendicular to the flow. This technique has been optimized for stratified types of flow and some slug flows, all without entrained bubbles, and yields void fractions with a measured precision in absolute value on the order of ±0.01 over the range from 0.05 to

392


0.95. As shown in this study, the technique is more difficult to apply to small diameter tubes than larger ones. Based on a statistical comparison, the drift flux model of Rouhani and Axelsson [15] as modified by Steiner [16] for horizontal flows is the best of the four models tested and accurately predicts the stratified flow void fraction data. Secondly, the Taitel and Dukler [18] stratified flow transition model for liquid height gives nearly as accurate results for calculating the void fraction. New dry angle measurements are also reported for an 8.0 mm sightglass tube and are accurate to about ±10.

Acknowledgment This investigation was supported by the Swiss National Science Foundation (FNS) contract number 21-57210.99.

References [1] T. Ursenbacher, L. Wojtan, J.R. Thome, Dynamic void fractions in stratified types of flow, part I: new optical measurement technique, Int. J. Multiphase Flow 31 (2004). [2] L. Wojtan, T. Ursenbacher, J.R. Thome, Dynamic void fractions in stratified types of flow part II: measurements for R-22 and R410a, Int. J. Multiphase Flow 31 (2004). [3] N. Kattan, J.R. Thome, D. Favrat, Flow boiling in horizontal tubes: part 1––development of a diabatic two-phase flow pattern map, J. Heat Transfer 120 (1998) 140–147. [4] N. Kattan, J.R. Thome, D. Favrat, Flow boiling in horizontal tubes: part 2––new heat transfer data for five refrigerants, J. Heat Transfer 120 (1998) 148–155. [5] N. Kattan, J.R. Thome, D. Favrat, Flow boiling in horizontal tubes: part 3––development of a new heat transfer model based on the flow pattern, J. Heat Transfer 120 (1998) 156–165.

[6] J. El Hajal, J.R. Thome, A. Cavallini, Condensation in horizontal tubes, part 1: two-phase flow pattern map, Int. J. Heat Mass Transfer 46 (2003) 3349–3363. [7] J.R. Thome, J. El Hajal, A. Cavallini, Condensation in horizontal tubes, part 2: new heat transfer model based on flow regimes, Int. J. Heat Mass Transfer 46 (2003) 3365–3387. [8] J.R. Thome, G. Dunkel, L. Wojtan, Optical measurement of void fraction, in: 2nd Japanese–European Two-Phase Flow Group Meeting, Tsukuba, Japan, September 25–29, 2000. [9] E. Krepper, A.K. Krussenberg, H.M. Prasser, A. Schaffrath, High resolution void fraction measurements for the validation of flow maps and CFD codes, in: Two-phase Flow Modelling and Experimentation, Edizioni ETS, Pisa, 1999, pp. 1371–1378. [10] G. Costigan, P.B. Whalley, Slug flow regime identification from dynamic void fraction measurements in vertical air–water flows, Int. J. Multiphase Flow 23 (1996) 263–282. [11] Cartellier, Measurements of gas phase characteristics using new monofiber optical probes and real-time signal processing, Nucl. Eng. Des. 184 (1996) 393–408. [12] A.A. Kendoush, Z.A. Sarkis, Void fraction measurement by X-ray absorption, Exp. Thermal Fluid Sci. 25 (2002) 615–621. [13] L.J. Xu, L.A. Xu, Gas/liquid two-phase flow regime identification by ultrasonic tomography, Flow Meas. Instrum. 8 (3/4) (1997) 145–155. [14] G.D. Harvel, K. Hori, K. Kawanishi, J.S. Chang, Cross-sectional void fraction distribution measurements in a vertical annulus twophase flow by high speed X-ray computed tomography and realtime neutron radiography techniques, Flow Meas. Instrum. 10 (1999) 259–266. [15] Z. Rouhani, E. Axelsson, Calculation of volume void fraction in the subcooled and quality region, Int. J. Heat Mass Transfer 13 (1970) 383–393. [16] D. Steiner, Heat transfer to boiling saturated liquids, VDIWarmeatlas (VDI Heat Atlas), Editor: Verein Deutscher Ingenieure, VDI-Gessellschaft Verfahrenstechnik und Chemie-ingenieurwesen (GCV), Translator: J.W. Fullarton, Dusseldorf, 1993. [17] S.M. Zivi, Estimation of steady-state steam void-fraction by means of the principle of minimum entropy generation, J. Heat Transfer 86 (1964) 247–252. [18] Y. Taitel, A.E. Dukler, A model for predicting flow regime transitions in horizontal and near horizontal gas–liquid flow, AIChE J. 22 (1976) 47–55.