Measurement System of Bubbly Flow Using Ultrasonic Velocity Profile Monitor and Video Data Processing Unit

Transcription

1 Journal of Nuclear Science and Technology SSN: (Print) (Online) Journal homepage: Measurement System of Bubbly Flow Using Ultrasonic Velocity Profile Monitor and Video Data Processing Unit Masanori ARTOM, Shirong ZHOU, Makoto NAKAJMA, Yasushi TAKEDA, Michitsugu MOR & Yuzuru YOSHOKA To cite this article: Masanori ARTOM, Shirong ZHOU, Makoto NAKAJMA, Yasushi TAKEDA, Michitsugu MOR & Yuzuru YOSHOKA (1996) Measurement System of Bubbly Flow Using Ultrasonic Velocity Profile Monitor and Video Data Processing Unit, Journal of Nuclear Science and Technology, 33:12, , DO: / To link to this article: Published online: 15 Mar Submit your article to this journal Article views: 192 View related articles Citing articles: 13 View citing articles Full Terms & Conditions of access and use can be found at

2 Journal of NUCLEAR SCENCE and TECHNOLOGY, Vol. 33, No. 12, p (December 1996) Measurement System of Bubbly Flow Using Ultrasonic Velocity Profile Monitor and Video Data Processing Unit Masanori ARTOMt, Shirong ZHOU, Research Laboratory for Nuclear Reactors, Tokyo nstitute of Technology*' Makoto NAKAJMA, Matsubashi Heavy ndustry Co.*' Yasushi TAKEDA, Michitsugu MOR, Yuzuru YOSHOKA, Paul Schemer n~titut*~ Tokyo Electric Power CO.*~ Japan Atomic Power CO.*' (Received July 11, 1996), (Revised October 8, 1996) The authors have been developing a measurement system for bubbly flow in order to clarify its multidimensional flow Characteristics and to offer a data base to validate numerical codes for multi-dimensional two-phase flow. n this paper, the measurement system combining an ultrasonic velocity profile monitor with a video data processing unit is proposed, which can measure simultaneously velocity profiles in both gas and liquid phases, a void fraction profile for bubbly flow in a channel, and an average bubble diameter and void fraction. Furthermore, the proposed measurement system is applied to measure flow characteristics of a bubbly countercurrent flow in a vertical rectangular channel to verify its capability. KEY WORDS: two-phase flow, measurement system, multi-dimensional flow, bubbly flow, ultrasonic velocity profile monitor, video data pmcessing unit, velocity, void fraction, pmbability density function. NTRODUCTON Many concepts have been proposed for the next generation LWRs in which passive and simplified safety functions are actively introduced with the aim of enhancing the reliability of their safety features, maintainability and so on. However, the driving force with passive safety features functioned by the law of nature is much smaller than that with active ones. Consequently, it is necessary with regard to passive safety features to be able to simulate multi-dimensional characteristics even for the two-phase flow which can be regarded as one dimensional flow for active ones. The two-phase flow shows essentially multi-dimensional characteristics even in a simple channel. The safety analysis codes such as the TRAC and RELAP5 codes treat the flow basically as one dimensional flow and introduce multi-dimensional ~ ~~~ *' Ohokayama, Meguro-ku, Tokyo 152. *' Wadasaki-machi, Hyogo-ku, Kobe 652. *3 Wurenlingen and Villigen, CH-5232, Villigen, SWTZERLAND. *4 Uchisaiwai-cho, Chiyoda-ku, Tokyo 100. *5 Ohtemachi, Chiyoda-h, Tokyo 100. Corresponding author, Tel , Fax , maritomionr.titech,ac.jp convection effects in a macroscopic way due to a lack of a fundamental data base for establishing the model of multi-dimensional two-phase flow dynamics. Therefore, it is one of the important problems for two-phase flow analysis to establish analytical methods of multidimensional two-phase flow for an analytical verification of the effectiveness of passive safety features. Measurement methods of two-phase flow characteristics are classified into two types: contact and noncontact. The contact type sensor such as an electric probe is inserted in a channel and thus disturbs the flow, so that the measurement accuracy is still not satisfactory. There are measurement methods based on ultrasonic, capacitance, conductance, optics and radiation in non-contact type methods. Laser Doppler Anemometry (LDA) based on an optical technique is excellent particularly in space and time resolution. Sheng and rons(l) reviewed LDA techniques related to two-phase flow measurements. However, this method is not only limited for void fraction measurement and for opaque fluids but also requires a long time for measuring a spatial distribution of flow characteristics in a channel. Recently, various radiation techniques with X-ray, y-ray and neutron have been developed, but they can be used only 915

3 916 M. ARTOM et al. in limited laboratory scale because adequate shielding is necessary. The authors attempt to apply an Ultrasonic Velocity Profile Monitor (UVP) to measure multidimensional flow characteristics in bubbly flow. The two-phase flow measurement systems using ultrasonic techniques are classified into the following categories: (1) Doppler shift method Lynnworth(2), and Brimley and Char~g'~) measured the liquid velocity in two-phase flow with low void fractions. Laheyc4) modified this method to measure the velocity of gas bubbles. They measured the velocity of a single bubble but did not measure that of multiple bubbles. (2) Pulse echo method A pulse echo method is based on the reflection of an ultrasonic pulse at the gas-liquid interface and the distance from the interface is obtained by measuring the transit time of the pulse. Morala et ~ 1.(~) measured the position and the size of a single bubble by this method. Chang, chikawa and rons(') applied it to measure liquid film thickness of an annular mist flow and characterized the flow regime. (3) Transmission method A transmission method is based on the propagation time of an ultrasonic pulse to measure the liquid velocity or on the attenuation of an ultrasonic wave to measure bubble rising velocity and size. Morala et uz.(~) investigated the bubble rising velocity and size in bubbly flows for a single bubble and a low void fraction using this method. Recently, an ultrasonic Doppler method for velocity profile measurement has been developed for liquid flow measurements(7). This method is advantageous over other methods described above, being able to obtain an instantaneous velocity profile of liquid flow. t has been approved that this method is a powerful tool in flow mea- surement in the following ways: t measures a velocity profile instantaneously so that velocity field can be measured in space and time domain@). Flow mapping can be efficiently performed since it measures velocity distribution on a line(@. However, since bubble diameters and void fractions cannot be measured through the UVP, the authors attempted to introduce the Video Data Processing Unit (VDP) to measure them. n this work, a measurement system combining an UVP with a VDP has been developed, which can measure simultaneously the multi-dimensional flow characteristics of bubbly flow such as velocity profiles of both gas and liquid phases and a void fraction profile in a channel, an average bubble diameter and an average void fraction. The present measurement system is applied to a bubbly countercurrent flow in a vertical rectangular channel to verify its capability.. MEASUREMENT METHOD ULTRASONC VELOCTY PROFLE MONTOR OF 1. Experimental Apparatus Figure 1 shows a schematic diagram of an experimental apparatus, which is composed of a vertical rectangular channel, an upper tank, a pump, a lower feedwater tank, a subcooler and an air supply system. Air and water were used as working fluids. The measurement system consisted of the UVP, the VDP and a personal computer to record and treat data. Water was fed into the upper tank and flowed downward in the vertical rectangular channel of 10 x loox 500 mm made of Plexiglas as shown in Fig. 2. The water level in the upper tank was kept constant with an overflow nozzle which was connected to the lower feedwater tank. The flow rate was regulated by a flow control valve and measured by an orifice flow meter both of which were attached at the downstream end of the test section and the water level.. Upper tank Fig. 1 A schematic diagram of experimental apparatus JOURNAL OF NUCLEAR SCENCE AND TECHNOLOGY

4 ~ ~ Measurement System of Bubbly Flow Using UVP and VDP us : UlmOniC Measwing line A- E : Measwing points Fig. 2 Test section in the upper tank. Micro particles of nylon powder were suspended in water to reflect ultrasonic pulses. Water temperature is kept constant by a subcooler. The air supply system consisted of a compressor and a pressure regulation valve. Bubbles were injected from three needles located near the bottom of the channel. The air flow rate was measured by a float flowmeter and regulated by another flow control valve. An ultrasonic transducer was installed on the outside surface of the front wall of the channel and a gap between the transducer and the wall was filled with a jelly to prevent a reflection of ultrasonic pulses on the wall surface, as shown in Fig. 2. After both air and water flow rates were set up at the desired values, a velocity profile along a measured line was measured by the UVP. The hydrostatic head was simultaneously measured as a pressure drop between the pressure taps installed on the side wall using a differential pressure transducer to get an averaged void fraction. The measurement error of the average void fraction is estimated as f5%. An outline of video camera equipment is shown in Fig. 3, which consists of an 8 mm video camera, a light source and a translucent sheet to unify the luminance brightness. The speed, diaphragm and gain of the video camera can be manually regulated and a speed of 60 flames per second can be obtained. After videotaping, the video digital data were recorded in a personal computer through an image converter. The picture elements are 640x240 dots, the color is monotone, and the brightness resolution is Measurement Principle The working principle of the UVP is to use the echo of ultrasonic pulses reflected by microparticles suspended in the fluid. Since the detailed information of its measurement principle was reported by Taked~(~)( ), the outline of the measurement principle of the UVP shown in Fig. 4 is explained in this paper. An ultrasonic transducer takes roles of both emitting ultrasonic pulses and receiving the echoes, that is, the backscattered ultrasound is received for a time interval between two emissions. The position information, x, is obtained from the time lapse, r, from the emission to the reception of the echo: x = cr/2, (1) where c is sound speed in the fluid. An instantaneous local velocity, uuvp(xi), as a component in the ultrasonic beam direction, is derived from the instantaneous Doppler shift frequency, fd, in the echo: UUVP = cfd/2f (2) where f is the basic ultrasonic frequency. The velocity resolution is given by GUUVP = UUVPmax/l28. (3) The UVP specification used in this work is tabulated in Table 1. US US transducer US burst 8. v US : Ultrasonic Wall X Measuring line Reflection from wall C : mage convelter PC : Pelsod computer Translucent sheet t vx US echo (Reflection from particles) video CBmela Test section Fig. 3 An outline of video camera equipment Fig. 4 Measurement principle of the Ultrasonic Velocity Profile Monitor( ) X VOL. 33, NO. 12, DECEMBER 1996

5 918 M. ARTOM et al. Table 1 The specification of the Ultrasonic Velocity Profile Monitor Basic ultrasonic frequency 4MHz Maximum measurable depth 758 mm(variab1e) Minimum spatial resolution 0.74 mm Maximum measurable velocity 0.75 m/s(variable) Velocity resolution 0.75 mm/s(variable) Measurement points 128 The number of profiles 1, Data Processing Method Since the sound speed of the longitudinal wave is the most fundamental parameter for this method, it is not possible to treat a two-phase medium as a homogeneous single phase medium, because a sound wave experiences multiple reflection among bubbles and its path returning to the transducer cannot be straight. t is however possible to obtain velocity profiles of liquid phase until the position of the nearest bubble from the transducer. Therefore, the authors attempted to derive information from each individual profiles by analyzing their shapes. The authors collected 9,216 (1,024~9) velocity profiles per one experimental condition and treated them statistically and it takes about 30 minutes to get them. The measured velocity profile is expressed by a location number, i, a profile number, j, and a velocity value, k: k = u[i,j]. (4) A position, y(i), is determined from the wall location, il and i2 as (5) where W is a channel width of 10mm. The measured velocity, which is a component in an ultrasonic beam direction, uuvp, is determined by UUVP = ~ Au, (6) where Au is a conversion factor from Doppler unit to velocity. The velocity in the flow direction, u, is then given by u = uuvp/cose, (7) where 8 is setting angle of transducer to the flow direction. The measurement error of velocity is estimated as f1.05 mm/s which is the velocity measurement resolution of the UVP. Since the velocity information is derived from Doppler shift frequency, no data must be available for the wall which is at rest. t is therefore possible to identify the wall position in the profile themselves. n practice, the wall position is defined as a location where the probability of data existence is 50% in liquid single phase flow. The measurement error of the wall position is estimated as f0.1mm and the measurement error of the location is estimated as f0.6 mm. The probability of data existence, Pl(y) is defined by N j=l Pl(4 = c fl(i,j) N, fi(i,j) = 1 for V[i,j] # 0, =O for V[i,j] =O, where N is the number of total profiles. A profile of the probability of data existence, P,(y), is obtained by converting the location number, i, into the real position, Y. Let us consider a velocity probability function: N j=1 1" j=1 P2(i,k) = Cf2(i,j) cfl(i,j), fz(i,j) = 1 for V[i,j] = k, = 0 for V[i,j] # k. A probability density function, Pu(y, u), is obtained by normalizing Pz(i,j) and by converting the location number, i, into the real position. A probability density function includes the velocity information of both phases. Assuming that each probability density function of both phases can be expressed by a normal distribution, the probability density function of mixture velocity is given by Pu (!, = &(Y)N[aG(Y), u& (911 (u> +( - &(Y))"aL(Y), 4(9)l(u), (10) where tic, ii~, UG and UL are average velocities and standard deviations of both phases respectively, E(Y) is the probability of bubble existence, which is expressed as 1 (u- ti)2 "a,u2](u) = -&Fexp[-T These five variables, UG, ti^, UG, OL and E, are calculated numerically by the least squares method. Since the ultrasonic pulse is reflected at the interface as long as a bubble exists, the bubble velocity can be always detected as an interface velocity. On the other hand, the ultrasonic wave is not reflected in water where a microparticle does not exist. As a result, water velocity is not always measured in the profile. Therefore, it is necessary to revise the probability of bubble existence as follows: n(y) is called the probability of bubble data existence in this work. An example of the probability density function obtained from the UVP data is representatively shown in Fig. 5. t is difficult to derive the genuine information under high void fraction conditions because the multiple reflection of ultrasonic pulse is induced by bubbles. Moreover, very little information on bubble velocities can be obtained at very low void fractions. To solve these problems, several data processing programs were (8) (9) JOURNAL OF NUCLEAR SCENCE AND TECHNOLOGY

6 Measurement System of Bubbly Flow Using UVP and VDP 919 = t i.g b m n e - n i i Velocity (mls) Fig. 5 A typical experimental result of the probability density function measured by the UVP developed in this work, for which a flow chart is shown in Fig. 6. These programs are described below: Since bubbly countercurrent flows are dealt with in this work, positive velocity data means bubble upflow velocity and negative velocity data does water downward one. The PROCl code is a program which selects positive velocity data before the position where the maximum bubble velocity appears and cuts off them behind the position in order to eliminate wrong data induced by a multiple reflection under conditions of high void fraction. The PROC2 code is another program to pick out only the maximum bubble velocity in one profile as bubble velocity data and cuts off other positive velocity data. The PROC3 code is another program to select only profiles including bubble velocities and is effective under conditions of very low void fraction. t is clarified that the both fig and ti1 in the probability density function velocity does not change even if the original data are VELQClTYcode FREQUENCYcode PROCl,PROcZ, PROC3 codes treated with the PROC1, PROC2 and PROC3 codes. Therefore, these programs were used only to get fig and cg in the probability density function of bubble velocity. The FREQUENCY code is a program which analyzes pulse height of measured velocities to give a velocity probability distribution at each point from the results for a measured profile. Since zero velocity cannot be distinguished from the data when the reflection wave is not received, the probability of the velocity number of 0 is substituted by averaging the values for the velocity numbers of -1 and 1. The SUM code sums every probability distribution calculated by the FREQUENCY code through one experimental condition. The VELOC- TY code picks up a probability distribution at one point from the data obtained by the SUM code and converts the location and velocity numbers into the real position and velocity as shown in Fig. 5. The SEPARATE code is used to calculate the five variables in the probability density function. Since the five variables cannot be solved analytically, iterating calculation is adopted in this code that the square sum of errors between its measured probability density function and Eq.(lO). Figure 7 compares a typical probability density function of mixture velocities calculated by the abovementioned procedure with experimental results. n this figure, open circles mean the results measured by the UVP and the line indicates the calculated result. U. MEASUREMENT METHOD OF VDEO DATA PROCESSNG UNT Frequency profile of velocity number Probability density function profile of mixture velocities Probability profile of data existence A typical picture recorded in the computer is shown in Fig. 8. n this work, it is assumed to get an average bubble diameter and void fraction that bubbles are not overlapped in the measured line and that the horizontal cross section of each bubble is a circle. The video data processing method is discussed below: (1) Since the brightness is not uniform in the whole picture even though the translucent sheet was adopted, the picture in a liquid single phase flow is taken as the standard one. SEPARATE code Time-averaged velocities of both phases. Bubble data existence probability and Turbulent intensity at each uoinr Fig. 6 A flow chart of data processing programs Velocity (m/s) Fig. 7 A typical analytical result of the probability density function using SEPARATE code VOL. 33, NO. 12, DECEMBER 1996

7 920 M. ARTOM et al. Fig. 8 A typical picture taken with a video camera (2) The differential picture of a bubbly flow was made by subtracting its recorded pictures directly from the standard one to eliminate the effect of the brightness in the picture. (3) The brightness in the differential picture at the interface between water and a bubble is weak due to the scattering and refraction. As a result, the brightness value is 0 for a liquid phase and a certain positive value for the interface. Using this feature, the data of the differential picture is converted into two-valued variables, 0 and 1, by a threshold value: 0 means a liquid phase and 1 does the interface. (4) Variables at every point in the inside surrounded by the interface are replaced with 1. Figure 9 demonstrates a typical two-valued picture obtained by this method. The number of bubbles in the twovalued picture is counted. (5) Each point in the picture is converted into the real coordinates based on both vertical and horizontal measures. (6) The center and diameter of each bubble cross section, Ci(z) and Di(z), on a horizontal plane are analyzed on the basis of the aforesaid assumption, and a void fraction, a(z), in the horizontal plane is ob- t ained bv Fig. 9 A typical two-valued picture where A is the cross section area of the channel, loox10mm and N is the number of bubbles in the two-valued picture. (7) The average void fraction is calculated by integrating a(.) vertically. The volume of the whole bubbles in the picture can be obtained from the average void fraction, and the average bubble diameter can also be obtained from the number of bubbles and the bubble volume. The brightness at the interface changes continuously due to the refraction and scattering of light. Therefore, we investigated the effect of the threshold brightness on the number of bubbles and a typical result is shown in Fig. lo(a). t is seen from the figure that there is a plateau offset against the threshold brightness value from 175 to 219 where the number of bubbles is not changed. Next, the effects of the threshold brightness on the average bubble diameter and void fraction were investigated and the results are shown in Figs. 10(b) and (c). t can be seen from these figures that the threshold brightness influences evaluation of the average bubble diameter and c....- F L a, P Average void fraction measured Threshold brightness (a) The number of bubbles Fig. Threshold brightness Threshold brightness (b) The average bubble diameter (c) The average void fraction 10 The effect of the threshold brightness JOURNAL OF NUCLEAR SCENCE AND TECHNOLOGY

8 (15) Measurement System of Bubbly Flow Using UVP and VDP 921 void fraction even though it does not affect the number of bubbles. There is no universal way how to determine the threshold brightness theoretically for the video data processing. The upper and lower brightness limits in the region where the number of bubbles is constant does not change for other experimental conditions in our experimental apparatus. Therefore, the threshold brightness of 197 is adopted in this work which is the median of the upper and lower brightness limits. The average void fractions obtained by measuring the hydrostatic heads are added in Fig. lo(c). t is noted that the measurement accuracies of the average bubble diameter and void fraction are &8% and to in the present method and that higher resolution of the VDP and its calibration can enhance their accuracies. lv. MEASURED RESULTS The developed measurement system was applied to bubbly countercurrent flows in a vertical rectangular channel. Figures ll(a) and (b) shows typical measurements of velocity profiles in both phases, UG and UL, and a probability profile of bubble data existence, n(y). Ve- locities of both phases are not zero on the wall because the ultrasonic pulse is emitted at an angle with respect to the channel wall and its diameter of 5 mm, which thus induces meaningful error of velocity measurements near the wall. However, this uncertainty is not a feature for two-phase flow measurement but appears for the velocity profiles measured for a single phase flow with the UVP. The probability of bubble data existence means that a bubble exists in an ultrasonic pulse path when the pulse is emitted, and is closely related to the void fraction. The bubble size, position and configuration cannot be known directly from UVP measurements. Supposing that the bubble size and configuration are at random and that they are statistically uniform at the whole points in the channel, the conversion factor, which relates the probability of bubble data existence to the void fraction, can be obtained. The following procedures are considered to get the conversion factor. The average volumetric flux of bubble <j~> is <jg>= ~ LjGdA - LQUGdA - A ' (13) A where j~ is the local volumetric flux of bubble and A the cross section area of the channel. Assuming that the local void fraction is proportional to the local probability of bubble data existence and that the proportional constant (the conversion factor), k, is uniform in the channel since it is dependent on bubble size and configuration, -0.21,,,,,,,,, jo= m/s jl=-0.06mls Distance from a wall (mm) (a) Velocities of both phases The average void fraction is expressed by k L nda <a>= ~ A * f air flow rate is known, transforming Eq.(14), jg= m/s jl=-0.06m/s JA Then, local void fraction, a(y), is given by 4Y) = qy). (17) The average void fraction, <a>, is also obtained by Distance from a wall (mm) (b) Bubble data existence probability Fig. 11 Typical profiles of mixture velocities and bubble data existence probability measured by the UVP <Cr>=k <K>. (18) Moreover, Fig. 12 shows a comparison of the average void fractions calculated by this procedure with the ones obtained from hydrostatic head measurements. The measurement accuracy in this procedure is within 25% error. Next, let us consider another procedure to get the conversion factor between the probability of bubble data existence and the void fraction using the average void VOL. 33, NO. 12, DECEMBER 1996

9 922 M. ARTOM et al. 1 v2 = yt(du + Db)2L. (22) 0.06 r/ 1 The conversion factor k2 is 0 jl=-0.06m/s o jl=-0.12m/s i a AP Fig. 12 Comparison of average void factions obtained by the VDP with those by the measurement of hydrostatic head fraction measured by the VDP. Supposing that the whole bubble exists in an ultrasonic pulse path when the pulse is emitted, the whole measurement volume, Vl, is expressed by 1 4 v, = - T ~ : ~, (19) where D, is the pulse diameter and L the UVP measurement length. The bubble volume is 1 vb = -TDZ, 6 where Db is the average bubble diameter. The conversion factor, kl, can be given by f the vertical surface of a bubble against the pulse direction exists in the pulse path, the ultrasonic transducer may receive the reflection wave of the pulse. Hence, we consider the case where a half of a bubble exists in the pulse path. n this case, the UVP measurement volume, V2, can be expressed by The conversion factors obtained from experiments are shown in Fig. 13 in reference to average bubble diameters. kl and kp are added in the figure. t is seen from the figure that kv, is located in the region of k2<kv<lc1. Although the conversion factor cannot be evaluated exactly because of a wide region between kl and k2, it is clarified that the factor is dependent on the average bubble diameter. Since the bubble is not spherical, the factor may also be influenced by its configuration. However, its empirical correlation will be established by analyzing numerous data. Moreover, it seems that the conversion factor is evaluated from the UVP data if each data profile is analyzed in detail and the pattern recognition of the reflection features is established. These subjects are future problems. The conversion factor can also be calculated by the average void fraction obtained from the hydrostatic head measurement and by the average probability of bubble data existence. Figure 14 demonstrates a typical void fraction profile converted from the probability profile of bubble data existence in this way. As mentioned above, the average void fraction in the flow cross section is estimated as f5%. With regard to local void fraction, although the error induced by the assumption that the conversion factor between the local void fraction and the local probability of bubble data existence is uniform in the flow cross section cannot be evaluated exactly, the error of the local void fraction may be estimated as &lo%. V. CONCLUSONS A measurement system combining an Ultrasonic Velocity Profile Monitor with a Video Data Processing Unit has been developed and proposed to measure multi i c 0 ~ x Average bubble diameter (Db) (mm) Fig. 13 The conversion factor from bubble data existence probability to void fraction Fig. 14 A typical profile of void fraction measured by UVP and the hydrostatic head JOURNAL OF NUCLEAR SCENCE AND TECHNOLOGY

10 Measurement System of Bubbly Flow Using UVP and VDP 923 dimensional two-phase flow characteristics. This system is applied to bubbly countercurrent flows in a vertical rectangular channel to verify its capability and the following insights are clarified: (1) Velocity profiles of both gas and liquid phases and a void fraction profile in the channel, an average bubble diameter and an average void fraction can be measured simultaneously with the proposed measurement system under conditions of low void fraction. The bubble velocity profile can be also obtained from the particle tracer method and this measurement method is our future work. (2) This system offers the probability density function of velocities in both phases. (3) The VDP measurement accuracies of the average bubble diameter and void fraction are within the error 333% and to , their accuracies can be enhanced by adopting higher resolution of the VDP and by calibrating it. The multi-dimensional characteristics of bubbly countercurrent flows measured by this proposed mea- surement system will be discussed in our next paper. This work was carried out at the Tokyo nstitute of Technology in collaboration with the Tokyo Electric Power Company, the Japan Atomic Power Company and the Paul Scherrer nstitut. -REFERENCES- Sheng, Y.Y., rons, G.A.: nt. J. Multiphase Flow, 17, (1991). Lynnworth, L.C.: Physical Acoustics, Academic Press, Vo1.14, (1980). Brimley, W.J.G., Chang, J.S.: Thermal-Hydraulics of CANDU Reactors, Chap.17, MES Press, Hamilton, (1980). Lahey, R.T.: Adv. Nucl. Sci. Technol., 13, (1981). Morala, E.C., et al.: Proc. 3rd Multi-Phase Flow and Heat Transfer nt. Symp., (1983). Chang, J.S., chikawa, Y., rons, G.: Measurements in Polyphase Flow, ASME Press, New York, 7-12 (1982). Takeda, Y.: Exp. Thermal Fluid Sci., 10, (1995). Takeda, Y., Fischer, W.E., Sakakibara, J.: Science, 263, (1994). Takeda, Y.: Nucl. Eng. Des., 126, (1990). VOL. 33, NO. 12, DECEMBER 1996