Master thesisInner flow structures of turbidity currents based on applied ultrasonic techniques

Download the full PDF ➡

Read the full paper ⬇

Master thesis

Inner flow structures of turbidity currents based on applied ultrasonic techniques

Jumpei HITOMI

Laboratory for Flow Control,

Division of Energy and Environmental Systems, Graduate School of Engineering,

Hokkaido University.

1st February 2019

Abstract

The flow behavior of experimental turbidity currents is evaluated quantitatively in order to reveal their long-distance propagation mechanism. Turbidity currents occurring in nature have complicated and large-scale of flow structures, so it is difficult to obtain sufficient knowledge about their flow behaviors from currents with a limited occurrence observed by a field research. Our research group, therefore, thinks that it is important to obtain basic knowledge that can be extended to actual phenomena from experimental turbidity currents. The test section is an inclined rectangle flume with 4,548 mm in length, 210 mm in height and 143 mm in width. The gate, which separates a heavier fluid and an ambient fluid, is positioned at roughly half length of the flume. It is revealed that the gravity currents generated by the lock-exchange method in such a high-volume release propagate with almost constant values of front velocity Uf. As the heavier fluids, three kinds of fluids are examined; quartz-suspended fluid, opalin-suspended fluid, and saline-dissolved fluids. The center diameters d50 of quartz particles and opalin particles are 12.2 and 18.9 m, respectively. Such small particles have been not used for UVP measurement basically, because the intensity of reflected waves from the particle is too small to be detected by an ultrasonic transducer. In this study, however, a supplemental experiment reveals that the volume fraction () range of quartz particles 0.1 ≤  ≤ 5% shows reasonable velocity distributions obtained by UVP with 4 MHz transducer. Total twenty-five cases of experiments are carried out changing the kind and initial density of the heavier fluids H. In the seven cases of them, two-layer turbidity currents are generated, and the experimental results are discussed. The details about the experimental facility, sediment materials, and initial conditions are summarized in chapter 2.

In chapter 3, analysis method is mentioned. To evaluate the flow behaviors of the gravity

currents quantitatively, it is essential to extend the use range of ultrasonic velocity profiler (UVP), which can be applied to opaque flows. In this study, the measurement by a pair of UVPs makes it possible to obtain horizontal and vertical velocity components (u and v) along the whole measurement line of the ultrasonic beam using time correction method based on Uf. Additionally, the particle concentration is estimated by the calculation using the spatio-temporal echo distribution. In case of Rayleigh scattering, the echo values vary with the particle concentration existing in a unit volume. In this study, the boundary condition at the bottom is set to solve the non-linear integral equation between the echo amplitude and the particle concentration. Besides, the scattering intensity is varying depending on the diameter of particles. Appling this characteristic to echo intensity distributions, it makes possible to detect interfaces in the cases of two-layer turbidity currents.

In chapter 4.1., the flow structures between the quartz-suspended currents and the saline density currents are compared. The initial density of all heavier fluids mentioned in this chapter is

H = 1008 kg/m3. Appling pattern matching method to sequential experimental images, it is revealed

that the Uf values of the turbidity currents are greater on average 19.0% than the density currents. The

wall shear stress W is calculated by the least square fitting method for the time-averaged u distributions. The W of the turbidity current is larger than the density current. Therefore, it is revealed that the turbidity current propagates with larger values of Uf than the density current, although the turbidity currents have larger friction coefficient Cf about wall shear stress on the bed. In the body region of the two types of currents, the maximum velocity of horizontal component (umax) is also larger in the turbidity current than the density current. On the other hand, both values of the height hm that

takes the maximum velocity umax and the thickness of the flows are smaller in case of the turbidity current. From these results, the existence of the suspended particles with heavier density than water suppresses the influence of the diffusion, which is predominant factor in the density currents, and helps the continuous supply of driving force due to the density difference.

In chapter 4.2., the turbulent flow structures of the quartz-suspended turbidity current as solid–liquid two–phase flow is discussed. The momentum conservation equation based on two-fluid model is applied to the distributions of the body region. The computed results indicate that the viscous stress and Reynolds shear stress among the five shear stress components are dominant like single– phase channel flows, but the distribution in the height direction of the shear stresses shows a different shape from that. The values of the viscous stress and Reynolds shear stress get canceled out in the outer region (hm < y < ht, where the u reaches zero), so it is revealed that the pressure gradient is working just in the inner region (0 < y < hm) of the current. Such a phenomenon shows a negative momentum transfer against the mean velocity gradient, which is working to keep a stable stratification. In chapter 4.3., the flow behaviors of several kinds of currents are discussed. These experimental cases show that the turbidity currents containing quartz particles propagates 38.5% faster than turbidity currents containing opalin particles. The analysis results about two-layer turbidity currents show that the kind of the particles suspended in the lower layer determines the front velocity Uf, when the bulk density H in initial condition is equal. In the two-layer turbidity currents composed of different kinds of particle-suspended fluid, it is confirmed that the lower layer is rolled up in the head region and the upper layer is covered with the fluid forming the lower layer. According to such a flow structure of the lower layer, not only the friction with the bed but also the interaction with the ambient fluid in the vicinity of the head region are dominated by the kind of lower layer, which means that the value of Uf can be almost estimated by the friction at the bottom and the interaction at the head region regardless the influence of the shear flow at the upper boundary. In addition to the comparison about Uf, the friction velocity u* is estimated from the distribution of the horizontal velocity component u in the body region. The relationship between the friction velocity u* and the maximum velocity umax in the body region is confirmed as u*/umax ≈ 0.114 in every experimental case. It is also

confirmed that the friction coefficient Cf =2(u*/umax)2 ≈ 0.0260 is obtained using the relationship.

Table of contents

Introduction 6
Experimental method 15
Development of analysis method 21
Results and discussions 39
Conclusion 86
Bibliography 89

Acknowledgements Appendix

‌Introduction
1. ‌Turbidity currentsGravity currents can be observed in many places, for instance in the seas, rivers, and so on. Turbidity currents, a kind of gravity currents driven by density difference between particle-suspended fluid and ambient fluid, are strongly related to the transportation or the deposition of fine particles. Consequently, they play important roles not only for a general understanding of global sediment transport process, but also for estimating the potential environmental hazards which they cause (reservoir sedimentation, trigger of Tsunami, cutting of submarine pipeline systems and cables, effluent dispersal and volcanic hazard (Middleton,1993)). Chamoun et al. (2016) suggested the problem about reservoir sedimentation and a solution for the efficient discharge by the venting. In order to settle the reservoir sedimentation problem, Oehy et al. (2010) evaluated the effect of inclined jet screen on turbidity currents. The result of their study indicates that in certain configurations turbidity currents can be partially stopped by the jet screen and the deposits downstream of the screen may be reduced up to a factor of two as compared with deposits of a free-flowing turbidity current. In addition, there is also the theory that the specific deposition formed by turbidity currents is related to the production or the melting process of fossil fuels such as methane hydrate. The reason why turbidity currents have been payed attention as an important subject is not limited in the field of civil engineering, and the interest for the long-distance propagation mechanism is attracting much attention with a view toward multi- phase fluid dynamics or Earth science.Actually, turbidity currents have been reported to propagate several thousand or hundred kilometers in the seas or rivers by field researches. For example, Azpiroz-Zabala et al. (2017) reported the measurement results by acoustic Doppler current profiler (ADCP) of actual turbidity currents occurring in Congo Canyon (see Table 1-1). In that research, turbidity currents are reported to have occurred six times in the period of seven months, and their thicknesses are from 48 to 77 m and the duration time of the currents is from 5.2 to 10.1 days. Additionally, Symons et al. (2017) also reported the field measurement results in Monterey Bay obtained by ADCP. In the study, the evolution of flow structure and composition are discussed as shown in Fig. 1-1.Contrary to such field researches, many laboratory experiments and numerical simulations have been conducted to collect detailed knowledge of the velocity and turbulent flow structures in the turbidity currents. Paker et al. (1986) suggested the self-acceleration mechanism of turbidity currents from numerical simulation. In the study, it is reported that the sediment entrainment from the bed resupplies potential energy, and it makes possible the acceleration and the long-distance propagation of turbidity currents. Cesare et al. (2001) established a novel numerical model for computational fluid dynamics of the turbidity current, which takes into account the interaction between the current and the deposited sediment. That three-dimensional (3D) numerical model can simulate the balance between deposition and erosion, and the currents provides good agreement with the turbidity currents in alaboratory flume as well as field measurements at the Luzzone Reservoir in swiss Alps. In the study by Gladstone et al. (2004), the multi-layer saline density currents were examined in the lock-exchange flume. It was revealed that the flow behaviors of those currents depend on a dimensionless densityratio between the layers * and dimensionless difference in the driving buoyancy B* (see Fig. 1-2);   g'*  U W  U , ( 1-1 )   g'andL W LB h g ' MB*  U  U U  U , ( 1-2 )U WU U L LU LB  B h g '  h g '  M  Mwhere U, L, and W denote the density of the upper layer, lower layer, and water, respectively. In the study by Longo et al. (2016), the saline density currents produced in a circular cross-section channel were focused on, and then they established a theoretical model which coincides well with the experimental results.‌Table 1-1 Summary of flow properties of the turbidity currents observed in Congo Canyons (Azpiroz-Zabala et al., 2017)
  ‌Fig. 1-1 Schematic of evolution of flow structure and composition of the turbidity current observed in Monterey Bay (Symons et al., 2017)
  ‌Fig. 1-2 Experimental images of the developing two-layer saline density current in case of * = 0.55 and B* = 0.36 (Gladstone et al., 2004)
2. ‌Ultrasonic measurement techniques and multi-phase flowsIn some laboratory experiments and industries, acoustic measurement techniques have been widely used, because they have strong advantages of (i) utility to apply to opaque fluids and inside the opaque materials (e.g. metal pipeline), (ii) non-invasive measurement, and (iii) wide measurable velocity range. Ultrasonic velocity profiler (UVP) (Takeda, 2012) is one of the most used devices in the field of fluid measurement, and UVP made it possible to measure the spatio-temporal velocity distributions along the ultrasonic beam, using frequency veering based on Doppler effects of ultrasonic waves. Along with the advance in ultrasonic measurement technique, the needs for ultrasonic measurement have also diversified and become complicated. In order to obtain such complicated flow fields, some advanced techniques have been established. One example is ultrasonic imaging velocimetry (UIV) or echo-PIV using array transducers (e.g. Poelma, 2017). In this method, two-dimensional echo image is generated by converting echo amplitude values received by each transducer element to gray scale values. Then, the cross-correlation method used in particle imaging velocimetry (PIV) is applied for two consecutive echo images, and vector fields can be obtained. In the study by Zheng et al. (2006), the results measured by UIV provided good agreement with the vector fields measured by optical-PIV (see Fig. 1-3).
  ‌Fig. 1-3 Validation of echo-PIV using a vertical flow: (a) B-mode particle image of the flow; (b) velocity field measured by echo-PIV; and (c) echo-PIV and optical PIV velocities along one radial line within the flow field (Zheng et al., 2006)Another one is phased-array transducer system (e.g. Kikura et al., 2016; Kang et al., 2016). In this method, adding a phase difference to the emitted ultrasonic waves from each element, spatial two- or three-dimensional echo images as well as measurement line can be generated. Murakawa et al. (2008) established unique ultrasonic measurement device using a dual-frequency Doppler transducer for bubbly flows, which makes it possible to measure both velocity distributions of tracer particles (liquid phase) and bubbles (gas phase) (see Fig. 1-4).
  ‌Fig. 1-4 Schematic image of measurement principle using ultrasonic multi-wave method (Murakawa et al., 2008)In addition to the contraption for the arrangement of ultrasonic transducers, another approach has been conducted, which is utilizing echo intensities information termed echo intensity method. Echo intensity, which is the strength of reflected ultrasonic waves scattered on interfaces between two media having different acoustic impedance, gives us beneficial information. In past studies, echo intensity obtained from UVP could be used to detect moving interfaces of air-water bubbly channel flow by Murai et al. (2010). Hitomi et al. (2017) used echo intensity of pulse repetition method to detect and monitor the air-oil-water three-layer pipe flows (see Fig. 1-5). Su et al. (2017) revealed the relationship between the attenuation coefficient and the phase fraction in oil-water two- phase pipe flows. Shi et al. (2019) established the method to characterize oil-gas-water three-phase flow using time-frequency decomposition. Dong et al. (2015) measured velocity distributions of oil- water two-phase flows using continuous ultrasonic waves. Additionally, echo intensity method was applied to obtain the profiles of suspended sediment concentration mentioned below and to measure the profiles of void fraction in bubbly flows (Murai et al., 2009).
  ‌Fig. 1-5 Samples of interface detection: (a) optical visualization, (b) echo intensity distribution,(c) phase distribution, and (d) Doppler velocity distribution (Hitomi et al., 2017)
3. ‌Final goal and objective of this studyAs mentioned above, the final goal of this study is to elucidate the long-distance propagation mechanism of turbidity currents occurring in nature under rivers, seas, and so on. In order to achieve this purpose, it is required to reveal the inner flow structures and the condition determining the flow behaviors of turbidity currents. The turbidity currents in nature, however, have huge flow structures, so it seems difficult to obtain sufficient and quantitative detailed data from field researches. Additionally, these have complicated flow structures due to turbulence and complex interaction caused between each particle. The turbidity currents show local mixture density fluctuation accompanying clusters and clouds of particles, which is hardly investigated by optical approaches and numerical simulations. Our research group, therefore, has ascribe experimental studies as a suitable option to take not only basic but also advanced knowledge to estimate the behavior of turbidity currents.To evaluate the flow behaviors of such experimental turbidity currents quantitatively, the measurement techniques, which can be applied to opaque flows with high resolutions, is required. Therefore, it is necessary to expand the utility of ultrasonic measurement techniques for solid–liquid two–phase flows. Based on the above discussions, the objective of this study is written below.
  - To elucidate the long-distance propagation mechanism of turbidity currents fromexperimental results by the lock-exchange technique
  - To expand the utility of ultrasonic measurement techniques, which can be applied to solid–liquid two–phase flows
  Experiments, analysis and discussions along the above objectives are described from the next chapter.
4. ‌Nomenclature
‌Experimental method
1. ‌Experimental facilitySeveral types of gravity currents were generated in a flume by means of the lock-exchange technique, which has been widely used to investigate unsteady currents in laboratory experiments (e.g. Theiler and Franca, 2016). Schematic diagram of the experimental facility is shown in Fig. 2-1 and the picture of the facility is shown in Fig. 2-2. The test section is an inclined rectangle flume with 4,548 mm in length, 210 mm in height and 143 mm in width. Surface on side and bottom walls of the flume walls is lubricated, and top boundary of the fluid layer is free surface. One of the walls is transparent whereas the other is coated with a non-reflective black film, producing a dark uniform and contrasting background for flow visualization. The region surrounded by a dotted square in Fig. 2-1 indicates the area for optical visualizations. A lock gate 1 made of plastic plate is positioned at roughly half length of the flume. As an initial condition, the left-side region of the flume separated by the gate 1 is occupied by a heavier fluid with bulk density H while the right-side region is filled with tap water with densityW as the ambient fluid. As experimental conditions of ambient fluid summarized in Table 2-1,temperature and density of tap water were almost constant, 8.0°C and 999.9 kg/m3, respectively. When the gate 1 is released, the heavier fluid invades under the ambient fluid with front velocity Uf according to the density difference between both fluids. The horizontal displacement from the gate in the right region is defined as x axis and y axis denotes the height from the bed. For one-layer turbidity currents, which are examined in laboratory experiments normally, one separation gate (gate 1 only) is enough as shown in Fig. 2-1(a). In this study, in addition to the one-layer currents, two-layer currents were examined using two separation gates (gate 1 and gate 2) as shown in Fig. 2-1(b) to observe the difference of flow behaviors depending on the initial conditions or variation of sediment particles. In cases of two-layer currents, the gate 2 was firstly released to generate density difference between heavier particle-suspended fluid in the tank A and lighter particle-suspended fluid in the tank B. Then, the gate 1 was barely released before the head of heavier particle-suspended fluid reached the gate 1, and two-layer currents propagated in the right-side region.
  ‌Fig. 2-1 Experimental facility of lock-exchange technique
  ‌Fig. 2-2 Picture of the experimental facility‌Table 2-1 Experimental conditions of the ambient fluids
2. ‌Sediment particles
  ‌Fig. 2-3 Pictures of sediment particles, left-side is the quartz particles and right-side is the opalin particles
  ‌Fig. 2-4 Particle images taken by electron microscopeAs suspended materials, two kinds of particles, quartz (Cario Bernasconi S.A., K-13) and oplain (Opalit AG Holderbank, Opalit), were examined (see Fig. 2-3), and their microscope photographs are shown in Fig. 2-4. Particle seize distributions of them measured by MasterSizer 3.0 (Malvern S. A., 2013) are shown in Fig. 2-5. In these distributions, a dispersion of the particle size distribution in case of the quartz particles is smaller than case of the opalin particles. The calculated center diameter and other physical properties mentioned in their specifications are summarized in Table 2-2. The calculated center diameters of quartz and opalin are 12.2 and 18.9 m, respectively. The quartz particles are almost composed of SiO2, while the opalin are composed of not only SiO2 but also Al2O3, Fe2O3, and so on. From these results, each particle-suspended fluid is expected to have different rheological characteristic and it would cause the change of their flow structures.
  ‌Fig. 2-5 Particle size distributions of (a) quartz particle and (b) opalin particles‌Table 2-2 Basic information about suspended particles
  As mentioned above, the specific weights of quartz and opalin particles are larger than 1.0, so the settling of the particles occurs in a still-water. Such a settling velocity vs is estimated by Stokes’ law written asd 2     gvs p p W18, ( 2-1 )Wwhere dp and g indicate the diameter of the particle and gravitational acceleration, respectively. The values of vs in case of quartz and opalin calculated based on each center diameter d50 are 0.0965 and0.244 mm/s, respectively. This law can be applied only for the particle Reynolds number Rep shown below is smaller than 2,Rep dpvs W . ( 2-2 )WAs the result of the calculation, the values of Rep for each particle were confirmed to be much smaller than 2, so the Stokes’ law works.
3. ‌Experimental casesTotally, twenty-five series of experiments using some kinds of different heavier fluids and initial conditions were performed, and experimental conditions about heavier fluids are summarized in Table 2-3. In the table, capitalized alphabet denotes the kind of heavier fluids, that is Q, S, and O show quartz-suspended fluid, saline-dissolved fluid, and opalin-suspended fluid, respectively. In Case 1–8, quartz-suspended turbidity currents, which are water containing quartz particles with 0.5% in volume fraction, were performed, and bulk density of heavier fluids H in those cases is 1008 kg/m3. In case 9and Case 10–12, heavier fluid contains quartz particles with 1.0% in volume fraction(H = 1016 kg/m3) and 2.0% (H = 1032 kg/m3), respectively. In case 13–15, saline density currents colored with white ink were examined for H fixed at 1008 kg/m3 in these three cases. In Case 16–18, opalin-suspended turbidity currents, which are water containing opalin particles, and H is differentfrom each other between 1008 to 1032 kg/m3. In Case 19–25, two–layer turbidity currents were performed. In Case 19–22, the density in the lower layer is 1032 kg/m3 and that in the upper layer is 1016 kg/m3. In Case 23–25 the density in the upper layer is 1032 kg/m3. The densities of each lower layer are 1016 kg/m3 in the Case 22 and 25, and 1032 kg/m3 in the Case 24.In this study, there are two types of the ultrasonic transducers arrangements as shown as Ⅰ orⅡ in Table 2-3, then the details of the arrangement and principle of velocity measurement by UVP are explained in next section.‌Table 2-3 Experimental conditions of heavier fluids
‌Development of analysis method
1. ‌Velocity measurementUltrasonic velocity profiler (UVP) (Takeda, 2012), which measures instantaneous profiles of velocity component along the propagation line of ultrasonic wave and is applicable to opaque fluids was adopted for the measurement of velocity profiles in the currents. In this method, using the frequency veering according to Doppler effect, spatio-temporal velocity distributions can be measured.‌
  Supplemental experiments by stirring flowAs mentioned above, quartz and opalin particles with order of 10 m in the central diameter are examined. This range of particle diameter is less than one-tenth the ultrasonic wavelength of central frequency 4 MHz in water, so Rayleigh scattering, which is almost isotropic scattering, occurs. Such small particles are not used for UVP measurement basically, because the intensity of reflected waves from a quartz particle is too small to be detected by ultrasonic transducers. In ordinary UVP measurements, it is considered that the adequate diameter of the tracer particles as reflector is appropriately quarter to half of the ultrasonic wave length. Fig. 3-1 shows the schematic diagram of those scattering patterns. In this part, therefore, a series of supplemental experiments was conducted by the stirring flows shown in Fig. 3-3 to elucidate the flow conditions and the volume fractions of the quartz particles that UVP can measure reasonable velocity distributions of particle-laden flows. The experiments were conducted in a cylindrical container with 100 mm in outer diameter, and 3 mm in thickness of lateral wall and 120 mm in height. The cylinder is made of acrylic resin and filled with test fluids containing the fine particles. The cylinder does not have a lid and thus top of fluid layer is free surface. The cylinder was mounted inside the water jacket to keep uniform temperature and allow transmission of ultrasonic wave from the outside of the cylinder. The flow was driven by a stirrer with 60 r.p.m. in the rotational speed, and a stirring bar dipped into the bottom of the cylinder. An ultrasonic transducer was fixed in the jacket with a horizontal displacement 10 mm from the center line. This off-axis measurement makes it possible to obtain the velocity component including not only radial but also azimuthal velocity component. The transducer was set at 25 mm from the bottom of the cylinder to avoid the blind of ultrasonic propagation due to the free surface. In Table 3-1, the setting parameters of UVP measurement for supplemental experiments are summarized. Each value of spatial resolution, number of cycles, and number of repetitions is same value in the measurements for the gravity currents in this study. The velocity range of the stirring flows is larger than that of the gravity currents, so temporal resolution is higher. The cylinder was filled with 500 mL of tap water and the quartz particleswere added based on each volume fraction . Total seven cases with different volume fractions,  =0, 0.01, 0.1, 1.0, 5.0, 10, and 15%, were conducted, and the sample images of the fluids for  = 0.01, 1.0, and 10% are shown in Fig. 3-2. As you can see, in cases of  > 1%, the test fluid becomes completely opaque, so the inner flow structures cannot be observed by optical approaches.
  ‌Fig. 3-1 Schematic diagram of scattering patterns reflected by each particle
  ‌Fig. 3-3 Schematic diagram of experimental setup for stirring flow‌Table 3-1 Setting parameters for supplemental experiments
  ‌Fig. 3-2 Sample images of test fluidsFrom these experiments, we tried to comprehend the trends of velocity profiles measured in this setup. Fig. 3-4 shows velocity distributions obtained from UVP in each experimental case. These velocities show the components of the measurement direction, that contain the azimuthal and radial velocity components. The measurements and the time-averaging were conducted for 40 s after the flows reached adequate developed states, so the corresponding number of the velocity profiles for time-averaging is 4000. In this figure, x axis donates the distance from the center of the cylinder and the ultrasonic measurement direction is from left to right side. In the cases of 0.1 ≤  ≤ 5%, reasonable velocity distributions can be observed. In this study, the velocity profiles measured with usual tracer particles (Mitsubishi Chemical, DIAION HP20SS, diameter 60–150 m, density 1020 kg/m3),which are indicated as orange dotted line in Fig. 3-4, is assumed as correct distribution, and “reasonable velocity distribution” means having good agreement with the correct distribution. The velocity profiles of 0.1 ≤  ≤ 5% have an accuracy that has a high cross-correlation value exceeding0.95 for the correct distribution. The cases of  = 10 and 15%, however, do not show the reasonablevelocity distributions. It seems that the influence of the scattering attenuation of ultrasonic waves occurring in cases of Rayleigh scattering becomes large and that prevents the ultrasonic transducer from detecting the echo signals. In addition, the case of  = 0.01% shows a notable result. The velocity values of this case drop irregularly near the center of the cylinder. This phenomenon implies that the amount of quartz particles is much fewer than wall side due to the centrifugal force. As known from the velocity distribution, high velocity values can be observed near the center. As the distance from the center becomes longer, the velocity values are getting smaller. From these distributions, the existence of a free vortex seems to be confirmed. The density of the quartz particles is about 2.65 times higher than that of water, so it is possible that the particles are blown from the center to near the wall sides by the centrifugal force.‌Fig. 3-4 Profiles of time-averaged value of Doppler velocity
  In addition to the velocity distributions, time-averaged echo amplitude profiles are also shown in Fig. 3-5. These echo amplitudes were calculated as the absolute values of difference from UVP echo intensity in case of background ( = 0%). As the volume fraction becomes larger in the cases of  = 0.01, 0.1, and 1%, echo amplitude values are also getting lager. In the case of0.01 <  < 1%, on the opposite side from the ultrasonic transducer, an extremely large value of echoamplitude is observed because of the reflection from the wall. In contrast, the reflected waves from the opposite wall disappears under the conditions  ≥ 5% due to the attenuation of ultrasonic waves, and the attenuation is getting larger depending on their volume fractions.
  ‌Fig. 3-5 Profiles of time-averaged value of echo amplitudeBasically, UVP measurements are applied to flows containing tracer particles with adequate size, density, and concentration to reflect ultrasonic waves. However, the concentrations of particles in these experiments are relatively larger in comparison with the flows which UVP measurements are applied to normally. The experimental results for 0.1 ≤  ≤ 5% show reasonable velocity distributions. This range of particle concentration, therefore, might be useful to detect integral echo signals obeying Rayleigh scattering. That is why UVP can obtain velocity distribution, although quartz diameter is much smaller than tracer particles (Hitomi et al., 2018).‌
  Velocity measurements for gravity currents
  ‌Fig. 3-6 Schematic image of arrangement of ultrasonic transducersAs mentioned above, there are two types of the ultrasonic transducer arrangements in this study. In more than half of the experiments, as shown in Fig. 3-6(Ⅰ), a pair of ultrasonic transducers was used for the measurement. Such a symmetric arrangement of two measurement lines has been used in some studies, because that makes it possible to obtain instantaneous profiles of both horizontal and vertical velocity components (u and v) at their crossing point using following equations (Jamshidnia and Takeda, 2010; Kitaura et al., 2010 (in Japanese)),u  y,t   , ( 3-1 )u1  y,t   u2  y,t 2sinv  y,t   . ( 3-2 )u1  y,t   u2  y,t 2cos‌In many cases of the previous studies, the relationship shown by these equations has been extended in the entire height direction to estimate multi-component or multidimensional flow field assuming that the tilted angle  is sufficiently small. In this study, two ultrasonic transducers with 4 MHz in basic‌frequency connected to each equipment of UVP model Duo (Met-Flow S.A., Switzerland) were installed with tilt angle of 25° from the y axis. Those two measurement lines of the transducers intersect at (xI, yI) = (1005 mm, 15 mm), where yI is the height expected roughly that the horizontal velocity takes the maximum value umax according to previous researches. For instance, in the study by Keller et al., (1999), the height hm taking velocity maximum within the body region of the saline density currents is reported to occur at hm/d ≈ 0.2 in the rectangular lock-exchange flume using Laser-Doppler anemometry, where d is the flow thickness of the currents. The start time of the two-UVP measurement was synchronized with the release of the gate 1, continuing sufficiently long measurement time to analyze flow structures from head to body of the currents.
  ‌Fig. 3-7 Schematic diagram of concept of the time correction method ()In addition, since the positions of two ultrasonic transducers in this arrangement have a certain displacement outside the intersection point, our research group suggests the time correction method of velocity distributions using front velocity Uf (Nomua et al., 2018a; Nomua et al., 2018b). Although the details about Uf will be described in another section, the currents in experimental cases of high-volume release like this study propagate with almost constant front velocity. The high-volume release means that the initial volume of the heavier fluids is relatively similar to that of ambient fluid in this study. The schematic diagram of this method is described in Fig. 3-7. Using this concept, Eq.( 3-1 ) and Eq.( 3-2 ) are modified as follows,‌u  y,t   , ( 3-3 )u1  y,t  t   u2  y,t  t 2sin‌v  y,t   , ( 3-4 )u1  y,t  t   u2  y,t  t 2coswheret   y  yI  tan . ( 3-5 )U fIn the remaining several experiments, in order to observe the temporal and spatial development of the currents, three ultrasonic transducers connected to one UVP equipment were arranged in parallel as shown in Fig. 3-6(Ⅱ), and the measurement by the multiplexer mode was performed. The positions of each transducer were x = 500, 1000, 1500 mm, respectively. The setting parameters for UVP measurement are specified in Table 3-2.‌Table 3-2 Setting parameters for UVP measurement
2. ‌Concentration profilesIn addition to the velocity profiles using frequency veering based on Doppler effects, echo amplitude distributions reflecting waves scattered by suspended particles give us beneficial information on particle concentrations. The size of sediment material used in this study is much smaller than ultrasonic wave length  (100 m) with 4 MHz frequency in water, that causes Rayleigh scattering. For such conditions, Thone and Hanes (2002) suggested the relationship between the mean-square amplitude of echo signal <V2> received by the ultrasonic transducer and mass concentration M in kg/m3 as following.M  4 a3 N , ( 3-6 )3 S Pwhere, aS denotes the equivalent radius of suspended particles (≈ d50/2) and N is the number of particles per unit volume.2 2 4 r  rM r dr 2   r 4r2   r  W 0 ‌M r   V   e  V   e   , ( 3-7 ) Ks Kt   Ks Kt where, W is the clear water attenuation which is a function of the temperature only expressed as (Fisher and Simmons, 1977)W  1015 55.9  2.37T  4.477 102T 2  3.48104T3 F2 , ( 3-8 )with T in Celsius degrees and F is the sound frequency in 1/s. Then  is sediment attenuation constant.  3 4aSp , ( 3-9 )where,  is the normalized total scattering cross-section, and it is described as (Thone and Hanes, 2002)1.1  4  0.18kaS 4  31  1.3kaS 2   4  0.18kaS 43, ( 3-10 )where k is the wave number of the sound in water (k = 2/,  are the wavelength of ultrasonic wave, respectively).Additionally, Ks is a function of the scattering properties of the suspended sediment, Kt is a constant for the ultrasound system, and (r) accounts for the effect of the oscillating backscattered signal from the spherical spreading in the near field. Furthermore, Pedocchi and Garicía (2012) applied this relationship to the echo distributions obtained from UVP. Their study indicated that the echo signal displayed by the UVP software is result of the passing signal through AD converter. That divides echo signals into voltage variation with 14 bits, and the measurable range is in ±2.5 V (2.5 V = 8191; - 2.5 V = -8191). The echo signals are amplified by the “gain factors” (as “gain start” and “gain end” in the software). To amplify the echo values, it is reported that time variable gain (TVG) is used and unamplified echo signal can be obtained following this equation using absolute gain values of start Gs and end Ge. The details of amplification by UVP software are written in Table 3-3 and Fig. 3-8.r rs   V  Vamp   Gs  re rs , ( 3-11 )G  G s  e where, Vamp is amplified echo signal, and rs and re are the minimum and maximum measurable distance from the transducer.
  ‌Table 3-3 Gain factors to amplify UVP echo signals (Oliver Mariette from Met-Flow, personal communication 2018)
  ‌Fig. 3-8 Gain factors for UVP echo signals monitored by the ultrasonic transducers with 2 or 4 MHz in basic frequency (Oliver Mariette from Met-Flow, personal communication 2018)V 2
  Here, a non-linearly is contained in the Eq.( 3-7 ), because the s is also a function of M. To overcome this problem, Lee and Hanes (1995) established an explicit method to invert the equation by substituting the concentration at an initial point MI. The procedure for the deformation of the formula is described below. Firstly, on the both sides of Eq.( 3-7 ) are taken by logarithm. r 2      r ln M r   ln V 2‌  e4r  2ln r   2ln   4lnWr   M r dr  . ( 3-12 ) Ks Kt     Ks Kt   0 Next, it takes the derivative with respect to r on both sides of the equation,dV 2
          dr            rdM r   V 2 
     dr     2   4W  M r  , ( 3-13 )M  r V 2    where (r) can be treated as constant in case of the measurement for far field from the ultrasonic transducer not a near field (Downing et al., 1995). Therefore, the second term on the right side of Eq.( 3-12 ) is handled as a constant, and it became zero by differentiating with respect to r.dV 2
   dM r  dr  2rV 2V 2
             dr          r 4W  M r   4M r 2. ( 3-14 )   using the general solution of Bernoulli differential equations, the flowing equation can be obtained.V 2 r2e4WrC  4  V 2 r2e4W rdrM r  ‌, ( 3-15 )where C denotes the integral constant. To set the boundary condition M(r = rI) = MI, finally it becomesr 2 V 2 e4W rM r  1  r2 V 2e4W rI  4 r 2 V 2e4W rdr‌. ( 3-16 )IrMI I I rWhen the ideal condition on lock-exchange technique, the MI is counted as zero, which means that the Eq.( 3-16 ) will diverge. It is required to set other boundary condition to overcome this problem. On the initial condition like this study, where the initial volume of heavier fluid is relatively similar to that of ambient fluid, the flow near the bed keeps containing the initial concentration as reported in (Theiler and France, 2016). That study shows the concentration profiles of colored density currents obtained by the image processing based on grayscale values, which was established by Nogueira et al. (2013). Applying this flow characteristic in case of turbidity currents where the high-volume release to this study, we rearranged Eq.( 3-15 ) by setting the boundary condition, where the concentration at the bottom is MB. In that case, the equation becomesr2 V 2 e4WrBBBM r  r   .rB B rBC  4Ir2 V 2 e4W rdr( 3-17 )Therefore, the integral constant C is determined asBr2 V 2C e4WrBB 4 rB r 2 V 2‌e4W rdr . ( 3-18 )MB rISubstituting Eq.( 3-18 ) into Eq.( 3-15 ), the following equation is obtained,M r  .V 2 r2e4Wrr2 V 2 e4WrBBBMB4 r V e dr  4rB2 2 4 rWrI V 2 r2e4W rdrrrI( 3-19 )r 2 V 2 e4WrThen, the equation is modified as,M r  1 r 2 V 2re4W rB  4  Br2 V 2e4Wrdr. ( 3-20 )MBIt takes finally the discrete form,B B rMn rn2 V 2Ne4Wrnn, ( 3-21 )1r2 V 2e4W rB  4r r2 V 2e4Wri  r2V 2e4W ri1 B   i i1 Mb bin1i i1 where n represents the position number of a certain measurement point when discretized.
3. ‌Pattern matching method and front velocityOutline motions of the fluids with the particles invading into the ambient fluid are monitored as successive images from t = t0 (Fig. 3-9), where t0 is an arbitrary elapsed time from the gate 1 release and the details about t0 will be mentioned later. The images were taken by the digital camera (Nikon, D5300) installed in front of the flume with 20 f.p.s. in flame rate for Case 1 and the shooting area is around xI as shown in Fig. 2-1. The frame rate was set to the same value as the sampling rate of the UVP measurements in cases of arrangement Ⅰ. In those images, some complicated flow structures can be observed near the upper boundary and the front part, which is caused by turbulent flow structure or interactions between the current and ambient fluid. In the study by Simpson and Britter (1979), tracing the shape of the vortices at the upper boundary, it was confirmed that the ratio of vortex structure to wavelength was the same as for Kelvin-Helmholtz vortices. In the image at t = t0 + 7.5 s of this experiment too, some transverse vortices can be seen clearly.
  ‌Fig. 3-9 Snapshots of quartz-suspended turbidity currents of Case 1 from t = t0 to t0 + 10.5 s
  ‌Fig. 3-10 Schematic diagram of pattern matching methodIn order to measure the front velocity Uf which is the advection speed of the head region of gravity currents, pattern matching method is applied to the successive experimental images. The schematic diagram of the pattern matching method is shown in Fig. 3-10. Firstly, a template image with 451 × 451 pixels was prepared from the image at t = t0, where t0 is a variable arbitrary elapsed time for each experimental case. Because the values of Uf are different from each experiment depending on the initial conditions, it is difficult to use a constant value as t0. In addition, it was also confirmed that the obtained result of Uf hardly depended on how to choose t0. In this study, therefore, the smallest value among the elapsed times of the image showing sufficient flow structures of the head region to make the template image with more than 451 × 451 pixels was used as t0 in each experiment. For every five images (every 250 ms), the cross-correlation value at each pixel was calculated, and just one pixel with the maximum cross-correlation value was defined as the head position of the current at that time. This operation was applied to the results of the experiments from Case 1–25. The results are shown in Fig. 3-11. In this graph, horizontal axis indicates the elapsed time (t) from t = t0 and vertical axis indicates the displacement of the computed head position from the initial position at t = t0. The results of quartz-suspended turbidity currents with 1008 kg/m3 in bulk density are summarized in Fig. 3-11(a). As you can see, a similar trend is confirmed in all experimental Cases 1–8, because the initial conditions are the same in these cases. Those results roughly ensure the reproducibility of the currents in this experimental facility. In Fig. 3-11(b), the symbols of square, diamond, and triangleindicate the results of quartz-suspended currents, saline density currents, and opalin-suspended currents, respectively. Additionally, the results of the two–layer turbidity currents in the Case 19–25 are shown in Fig. 3-11(c) by cross marks. As you can guess, in the two–layer turbidity currents, the variation of the gray scale values is larger than that of the one–layer turbidity currents, because two kinds layers showing different color from each other are mixed. Hence, in the cases of two–layer currents, it is confirmed that the variations in the head position detected by the pattern matching are slightly larger than in the cases of one–layer currents.
  ‌Fig. 3-11 Time variation of the head position of the gravity currents, (a) quartz-laden currents with H = 1008 kg/m3, (b) the remaining cases of the one–layer currents (quartz, saline, and opalin currents), and (c) two–layer currentsIn all cases, the displacement of the head is increasing in almost direct proportion to the tfdue to the lock-exchange experiments with high-volume release, that means the sufficient driving force keeps supplied by the density difference between the heavier fluids and the ambient fluid. Therefore, the currents propagate with almost constant front velocity Uf without decay in these experimental time ranges. Using this characteristic, the front velocity of each experimental case was calculated from the slope of a regression line estimated by the least-squared method, and the result is shown in Fig. 3-12. In this graph, the vertical axis denotes the calculated front velocity Uf and the horizontal axis denotes the initial bulk density of heavier fluids (H). In case of two–layer currents, the density is calculated with considering the volume difference between the tank A and the tank B. The corresponding values are 1023 kg/m³ in Case 19–22, 1012 kg/m³ in Case 23 and 24, and 1019 kg/m³ in Case 24. The symbol indicating each case is the same to that used in Fig. 3-11. In the past studies about saline density currents, the dimensionless velocity ( U * ) for lock-exchange method have been used to summarize thevalues of Uf, and it is defined asg'Hf‌U *  U f , ( 3-22 )where H is total fluid depth and g´ is reduced gravitational acceleration defined asg'  H  W g . ( 3-23 )WIn this study, however, the flume was inclined at 1.38° and the fluid depth varies with the distance from the gate. Here, the calculation was carried out using the fluid depth of H = 179 mm at x = xI asUfthe representative depth. Some values of * are reported in the previous studies about lock-exchange density currents, as 0.44 (Middlen, 1966), 0.46 (Keulegan, 1957; Barr, 1967), and 0.41 (Kneller et al., 1999). Therefore, the front velocity can be estimated using Eq.( 3-22 ) from each initial density of heavier fluids. In Fig. 3-12, the estimated front velocity is plotted by solid lines, the upperffone is the estimated velocity in case of U *  0.46 and the lower is in case of U *  0.41 . The valuesUfof experimental Uf and estimated by * have good agreement just in cases of quartz-suspendedcurrents and some two–layer currents of experimental Cases 19, 20, 23, and 24, where the lower fluids contain the quartz particles. Contrary to that, the saline density currents show a little bit lower values of Uf, that may be affected by the inclined experimental flume as a result of the balance between downward force of heavier fluid and upward force of counter flow by ambient fluid. Notable points are shown in the remaining experimental cases of opalin-suspended currents and two-layer currents whose lower layer contains opalin particles of the experimental Cases 21, 22, and 25. In those cases, the values of Uf are obviously lower than estimated values. These results suggest that the value of Uf cannot be estimated only by initial density of heavier fluids and depth, that seems to be attributed to the difference of rheological property between opalin-suspended fluids and quartz-suspended fluids. The details about it will be discussed later.
  ‌Fig. 3-12 Measured front velocity in experimental Case 1–25The spatio-temporal velocity distributions of the experimental Case 3 obtained from UVP are shown in Fig. 3-13, those distributions were measured at (a) x = 500, (b) 1000, and (c) 1500 mm, respectively. The axes denote the height from the bed (y) and the elapsed time from the release of the gate 1 (t). In the arrangement of transducers Ⅱ, the velocity components along the propagation line of ultrasonic wave (u) are obtained. The magnitude of u is shown as color contour. Comparing these three distributions, the head of the current can be observed clearly at the measurement line x = 1000 (b) and 1500 mm (c), but it cannot be done at x = 500 mm (a). Therefore, it is assumed that the current was developing between x = 500 to 1000 mm and reached stationary state before x = 1000 mm in this case. Here, in order to confirm the characteristic of Uf, the estimated arrival time t* is defined ast*  l ,xl U f( 3-24 )where, l is an arbitrary distance from the gate 1. The value of*txlrepresents the expected time ofarrival at each measurement line, assuming that the front velocity Uf calculated above will be constanteven if it reaches each measurement point. Each value of*txlat x = 500, 1000, and 1500 mm isshown in Fig. 3-13 as dotted line. From those results, it is confirmed that the turbidity current roughly reaches each measurement line at the estimated arrival time calculated from Uf. In this study, therefore, the calculated front velocity Uf in each experimental case cam be treated as constant value below.
  ‌Fig. 3-13 Spatio-temporal velocity distributions measured at (a) x = 500, (b) x = 1000, and (c)x = 1500 mm of experimental Case 3
‌Results and discussions
1. ‌Comparison of flow structures between quartz-suspended currents and saline density currentsIn this section, the differences of flow characteristics between quartz-suspended turbidity currents of Case 1–8 and saline density currents of Case 13–15 are discussed. The main driving force of the currents of this experimental facility is density difference between heavier fluids and ambient fluid, and all of the initial densities of heavier fluids discussed in this section are the same value ofH = 1008 kg/m3. In other words, it is the main target of this section to investigate how the flowbehavior of gravity currents vary with the presence or absence of quartz particles and to compare flow structures for the elucidation of the long-range propagation mechanism of the turbidity currents.‌
  Inner velocity structures of the currentsSpatio-temporal distributions of horizontal and vertical velocity components (u and v) obtained by two-UVP measurement are shown in Fig. 4-1, where (a) left and (b) right panels are from Case 1 (turbidity current) and Case 13 (density current). The axes denote the height from the bed (y) and the elapsed time from the release of the gate 1 (t). The V-shaped distributions seen in these graphs are the ultrasonic interference noise caused by the simultaneous measurement of the two UVPs. As the color scale corresponding to the velocity magnitude, horizontal velocity component u is dominant in the flow. Because lock-exchange flume is a closed system, counter flows of ambient fluid occur at the upper boundary of the currents. In the distributions of u, the increase and decrease of velocity due to the propagation of the currents are confirmed. Regarding the distribution of v, in spite of small velocity magnitude, strong rolling-up structure is observed near the head position in the turbidity current (20 < t < 25 s). Some pair of positive and negative values in v distributions near the upper boundary can indicate the existence of vortices attributed to Kelvin-Helmholtz (K-H) instability.
  ‌Fig. 4-1 Spatio-temporal distributions of horizontal (above) and vertical velocity component (below): (a) Case 1 (turbidity current) and (b) Case 13 (density current)‌
  Time variation of the height and velocity of the currentsThe timeline images made from snapshots are shown in Fig. 4-2, where (a) left and (b) right panels are from Case 1 (turbidity current) and Case 13 (density current) extracted at xI = 1005 mm (corresponding to the cross point of two ultrasonic measurement lines). In the front region of the turbidity current at the time about 20 s < t < 30 s, the edge of the head region can be observed clearly and the height of the current is larger than that in the body. Comparing with the image of turbidity current, the rolling-up structure at the head and shear waves observed in the body region seem to be smaller in case of the density current. In order to quantify the height of the two currents mentioned above, integral scale method (Ellison and Turner, 1959) is applied. Combining this method with u distributions obtained by UVP, the representative flow layer thickness h and layer-averaged velocity U can be calculated at each time, every 50 ms along the equations,htUh  0 u( y)dy , ( 4-1 )U 2h  ht u( y)2 dy , ( 4-2 )0where, ht is the height at which u(ht) = 0 near the upper boundary. Here, the spatio-temporal velocity distributions obtained by UVP contain measurement noises. To reduce the influence of the noises, median filter with 9  9 measurement points was performed for the spatio-temporal distributions. The time variations of h and U for case 1 and case 13 are shown in Fig. 4-3, where top panels show profilesof the ht (black square) and h (green dot). In the case 1, the obvious edge of the head cannot be observed in comparison with the timeline image, which means that both values of ht and h take the values of around y = 60 mm, although the height of the head region seems to reach y ≈ 90 mm in the timeline image. This result suggests that the complicated flow structure in the head region containing not only u but also relatively large magnitude of v component causes the underestimation of the flow thickness. From these two distributions of the flow layer thicknesses, it is confirmed that the turbidity current is thinner than the density current in whole region including the heads. The bottom panels in Fig. 4-3 show profiles of the layer-averaged velocity U (black diamond) and front velocity Uf (green dot line) as the reference value. In the vicinity of the front position, the magnitudes of U almost agree with Uf. On the other hand, the profiles of U always have larger values than Uf in the body region. These results suggest that in the two kinds of gravity currents the momentum is supplied from the wake flows, namely, the momentum is transported from the body region to the head region.
  ‌Fig. 4-2 Timeline image at the cross-point of two ultrasonic beams emitted from the transducers:(a) Case 1 (turbidity current) and (b) Case 8 (density current).
  ‌Fig. 4-3 Time variation of h and U calculated by the integral scale method of (a) Case 1 and (b) Case 13‌
  Flow structures in the body regionIn this section, the flow structures of the body, which can be treated as quasi-steady flow (Kneller and Buckee, 2000), is focused on.The profiles of the time-averaged u of Case 1 and Case 13 are shown in Fig. 4-4, with the averaging window of t = 35 to 70 s. There is no clear definition to separate the head and the body in the gravity currents. With judging from the h distributions from the integral scale method, t = 35 s was used as a separation point in this study, which means the height becomes lower than the head region after t ≈ 35 s. As mentioned in the discussion about integral scale, these two graphs indicate that the first zero-crossing point on y (ht), corresponding to the thickness of the currents, is smaller in the turbidity current than in the density current (in some previous studies, there are other definition of thethickness). Several time-averaged properties to describe the rough flow structures are shown in Table 4-1, where the overline (‾) means the time-averaged value from t = 35 to 70 s. In the Case 1, the velocity u reaches the maximum at y = hm = 14.6 mm and the corresponding velocity is umax = 79.2mm/s, and hm = 25.4 mm and umax = 73.2 mm/s in the Case 13. In case of particle-laden flows like turbidity currents, the density of suspended particles is larger than that of water normally, so the force direction works downward. Contrary to that, in the saline density currents, diffusion caused by the density difference is predominant, so the force direction is upward, that is opposite direction against the case of turbidity currents.
  ‌Fig. 4-4 Distributions of time-averaged horizontal velocity component u in the body region from t = 35 to 70 s for (a) Case 1 (turbidity current) and (b) Case 13 (density current), where red lines are the result of logarithmic fitting at 0 < y < hm and blue lines are the result of exponential fitting at hm < y < htPrevious researches reported that the u distributions of turbidity currents have a shape similar to a wall jet, where two regions are distinguished, namely, a wall region (y < hm) and a jet region (hm < y < ht). Altinakar et al. (1996) suggested the fitting method to the u distributions in the body for two regions. The fitting procedures are shown below.
  1. In the wall region, y < hm, by a logarithmic relation,u( y)  u * ln y y0, ( 4-3 )This equation is “law of the wall (e.g. Kida and Yanase, 1999 (in Japanese))” that predicts a logarithmic velocity profile in inner layer (y < hm), and the variables are u* and y0 which are friction velocity and zero-velocity roughness height, respectively. In this study, assuming a smooth bed, the value of 0.41 is substituted into the von Karman constant  (e.g. Zordan et al., 2018). The calculated u* and y0 are shown in Table 4-1. The fitting result based on this method is shown in Fig. 4-5, where the y axis denotes the semilogarithmic axis. The profiles of time-averaged horizontal velocity component u in the body region show a roughly linear distribution in a semilogarithmic graph at y < hm and the fitting result closely matches that. Besides, it is the big advantage of UVP that the measurement with high spatial resolution ofy = 0.67 mm can be performed in this case.
    ‌Fig. 4-5 Fitting results of the wall region at y < hm for (a) Case 1 (turbidity current) and (b) Case 13 (density current)
  2. In the jet region, y > hm, by a near-Gaussian relation,
    y  h m u( y)  umax expc  m  . ( 4-4 )  h  hm   This relation is fitted to the velocity profiles measured in the jet region namely outer region of the currents. The variables are m and c, and fitting results are shown in Table 4-2. The values of fitting variables in the Case 1 show good agreement with the values in the previous study (Altinakar et al., 1996). The fitting curve obtained from the variables are also shown in Fig. 4-4 by blue solid line. The values of m and c in the present study of the Case 1 coincide with the results of their research which used the same quartz particles.Using calculated u* in the wall region, the shear stress working near the bed W can be estimated based on the definition of friction velocity as, WHu*  . ( 4-5 )As the result of the estimation assuming logarithmic profiles, the wall shear stresses near the bed are0.086 Pa in Case 1 and 0.042 Pa in Case 13. It means that the turbidity current can propagate faster than the density current, in spite of the larger wall shear stress acting near the bed.‌Table 4-1 Time-averaged representative values to dictate the flow patterns in body of gravity current
  ‌Table 4-2 Fitting variable for the jet region
  ‌
  Comparison of fluctuation components in the body regionIn this section, the fluctuation components of the velocity as turbulent flows in the body region of the currents are focused on. Taishi et al. (2002) suggested the method to obtain fluctuation components and Reynolds stress using two measurement lines (like the arrangement of Ⅰ in this study). Using that concept, the root mean square values of the fluctuation components are calculated along following equation,u'RMSv'RMS  , ( 4-6 )u' 2
  u' 2  2u' u'  u' 21 1 2 24sin2 v' 2
  u' 2  2u' u'  u' 21 1 2 24cos2   , ( 4-7 )where, the averaging window of t = 35 to 70 s in this case too. The vertical distributions of the calculated those values are shown in Fig. 4-6. The distributions have notable characteristics. Firstly, u´RMS value is larger than v´RMS globally in both cases, and they take local maximum value at y = ht. Besides, in the upper region y ≳ ht, both values of u´RMS and v´RMS become larger than the lower region, because the interactions with ambient fluid, which include K-H instability mentioned above, attributed to the density or flow velocity differences occurs in the upper boundary. Such the interactions vary with space and time even in the body region treated as quasi-steady flow, which works to amplify theu´RMS and v´RMS. The outstanding point is vicinity of the bed (in the inner region: y < hm). In this region, the tendencies of t u´RMS and v´RMS distributions are obviously different each other case, that is to say, both values in the turbidity current are smaller than in the density current. From these results, it is predicted that the existence of quartz particles suppresses the flow fluctuations in the inner region of the turbidity current y < hm. In the distribution of the turbidity current, moreover, the values of u´RMSare getting larger dramatically from y ≳ hm. In fact, this positive peak in the u´RMS distribution is alsoconfirmed in the distribution of the density current at y ≈ 50 mm, but the spatial size and the magnitudeare much smaller than in the case of the turbidity current.
  ‌Fig. 4-6 Vertical profiles of the root mean square values of fluctuation components for (a) Case 1 (turbidity current) and (b) Case 13 (density current)‌
  Summary of the difference of flow structures between the turbidity currents and the density currentsThis part summarizes the comparison results of the flow behaviors between the quarts-suspended turbidity current and the saline density currents. A simple comparison table of some physical properties that determine flow structures of represent both currents is shown in Table 4-3, where only the magnitude relation of the respective values is represented by the size of circles ○. As shown in the table, the most essential difference of both currents occurs at the front velocity Uf. Using image processing, it was revealed that the Uf values of the turbidity currents in eight cases are greater on average 19.0% than the density currents in three cases. Therefore, the comparison of inner flow structures of those currents was conducted using ultrasonic techniques.‌Table 4-3 Measured physical properties determining the flow behaviors
  The time-averaged profiles of horizontal velocity component u in the body region show that in the case of the turbidity current, not only the Uf, but also the maximum velocity umax in that region is greater than the density current. On the other hand, the height hm that takes the maximum velocity is smaller in case of the turbidity current than the density current. Using these profiles, the shear stress W  u *2 working near the bed can be estimated based on a definition of the friction velocity u*obtained from the low of the wall, and the friction coefficient Cf is defined asCf  W ,1 u22( 4-8 )where,  shows the bulk density of the fluid, so H should be used in the present case. Here, considering the definition of W, the term of  is canceled out, and finally it is written as u * 2 .f uC  2  ( 4-9 )The horizontal velocity component should be substituted for u in the equation, but there is room for discussion as to which value is actually used such as Uf, umax, and time-averaged U. The calculated Cf values using each horizontal velocity value are summarized in Table 4-4. In all cases, the Cf values in case of the turbidity current is larger than the density current.‌Table 4-4 Friction coefficient using each u value
  The UVP measurement made it possible to obtain the representative flow layer thickness h and layer-averaged flow velocity U at each measurement time (every 50 ms in this study) using the integral scale method (Ellison and Turner, 1959), which can show the time variation of the h and U. In the body region of the currents, the h value, which roughly corresponds the height of the current, is smaller in the turbidity current than the density current.Considering those results, the estimated flow structures of both the currents are described in Fig. 4-7. In case of the density current, the diffusion of the saline heavier fluids may be dominant factor to form the shape of the currents due to the density difference with ambient fluid. Such a diffusion working upward direction in the body region may cause the enlargement of the density currents. As a result, the height, where the horizontal velocity takes maximum value, seems to become larger than the turbidity currents. Contrary to the case of density currents, the suspended particles within the turbidity currents have a heavier density than water as the ambient fluid in many cases, here the specific density of quartz particles used in this study is 2.65. Therefore, in the turbidity currents, not only diffusion of the particle-suspended fluids but also the settling of that particles occurs. It is considered that the height of the turbidity currents may be suppressed due to such the settling effect of the particles. From the vertical distributions of the root mean square values of velocity fluctuation components shown in Fig. 4-6, it was confirmed that the magnitude and the position of the positive peak of the u´RMS values are different from each case. Comparing the u´RMS distributions, the peak near the upper boundary can be observed in the both cases, but there are clear differences in the way the distribution at y < ht. In case of the density current, u´RMS distribution takes larger values as going downward in this region. On the other hand, in case on the turbidity current, the magnitude of the u´RMS shows large values in the range of hm < y < ht (outer region). From these results, it is predicted that the existence of the suspended particles would suppress the fluctuation effect near the bed. In addition, the interactions of particle–particle and particle–fluid are also caused, and it would work as a mixing effect for suspended particles, which suppress the settling of the particles. In the results of such a mixing effect, suspended particles can be re-supplied to near upper boundary, which generates the additional driving force, because the risen heavier particles get the potential energy again. Such a series of the processes may become a key factor to explain the reason why the turbidity currents can propagate faster than the density currents in spite of that the friction coefficient Cf of the turbidity current is larger than the density current.
  ‌Fig. 4-7 Estimated flow structures in (a) turbidity currents and (b) density currents
2. ‌Inner flow structures of turbidity currents as solid–liquid two–phase flow Firstly, the experimental images of the front motion at each elapsed time from the gate 1 release are shown in Fig. 4-8, where the current of the Case 1, which was composed of the water containing quartz particles with 0.5% in volume fraction, is focused on. These images were taken from the front of the current by a digital camera (GoPro, Hero4), which was mounted on the right edge of the flume and completely submerged in the fluid. In the figures, the complicated flow structure can be observed in the head region. In particular, it is characteristic that some large flow structures are wound up fromnear the bed, and the structures gradually becomes smaller as they gradually wrap upward. Such a change in the flow structure resembles the theory of the turbulent energy cascade in which it sequentially propagates from large-scale vortices to small-scale vortices (Kida and Yanase, 1999 (in Japanese)). As you can see, however, the information obtained from the optical images cannot enough explain the long-distance propagation mechanism on turbidity currents due to their opacity.In this section, therefore, the inner flow structures of the turbidity current as a solid–liquid two–phase flow are discussed using velocity and echo profiles obtained by UVP. The analysis was carried out using the experimental results of the quartz-laden turbidity current of Case 1 discussed in the last section.
  ‌Fig. 4-8 Snapshots of front motion at each elapsed time from the release of the gate 1 (turbidity current in the Case 1 containing quartz particles with 0.5% in volume fraction)‌
  Flow field of the currentThe use of a pair of UVPs makes it possible to obtain both velocity components of horizontal (u) and vertical v. Additionally, it was revealed that the currents propagate with almost constant front velocity Uf under the condition of high-volume release, which means that the amounts of the heavier fluid and the ambient fluid are almost same on the initial condition like this study. Therefore, although UVP measures velocity profiles at fixed line, it seems possible to reconstruct a flow field with space-space using Taylor’s frozen fluid hypothesis. The concrete procedure is as follows.Firstly, the front velocity Uf is calculated and the value is 49.9 mm/s in the Case 1. As shown in Fig. 4-1, the UVP measurement was continued for 70 s. Assuming that the turbidity current propagated at a constant front velocity Uf from the release of the gate 1, the head reaches about 3500 mm at t = 70 s. Utilizing such a property, the converted travel distance X at t = 70 s was calculated using the following definition,X  U f 70  t  t* x  , ( 4-10 ) x I where t* indicates the estimated arrival time at x = xI defined in the previous section. Because the velocity components could not be measured by a pair of UVP in the region t < t*, it was modified in the elapsed time t in advance. Using the converted travel distance X, the spatio-temporal velocity distribution obtained by UVP can be treated as a spatio-spatio distribution. The obtained vector field combining the distributions of u and v component are shown in Fig. 4-9. In the figures, the height from the bed y and the converted travel distance X are used for the axis, and the flow structures obtained from the optical timeline image and the velocity distributions are shown in Fig. 4-9(a) and (b), respectively. Regarding the vector field, from the aspect of visibility, the decimated result using the averaged value of 10 profiles in the X direction and 5 profiles in the y direction is shown. Comparing these two figures, the head region showing higher interface between the current and the ambient fluid can be observed around 300 < X < 3500 mm. In addition, a large rotating structure can be observed around y = 60 mm, which is expected to be caused by the shear with the ambient fluid. Paying attention to the body region, in the vector field, the fluctuating interfaces are also observed around y = 50 mm, and the height roughly corresponds to the zero-velocity height ht (as discussed in the last section, the time-averaged value in 35 s < t < 70 s is 50.7 mm). The distributions of vorticity calculated from the vector field are shown in Fig. 4-9(c). Normally, the vorticity is represented by the spatial differential of the velocity components shown asz  v  u . ( 4-11 )x yIt takes the discrete formz  v  u . ( 4-12 )x yIn this study, for the discrete width ∆x in the x direction, the following value ∆X was used with UfX  U f  tUVP , ( 4-13 )where tUVP is the temporal resolution of UVP measurement and the value in this case was 50 ms, so the ∆X denotes the value of 2.50 mm. As same to the discussion on vector distributions, in the vorticity distribution, relatively large positive values are observed near the interface. In the head region, the negative values also appear and that is caused by the complicated flow structures containing not only horizontal velocity u but also vertical velocity v. Contrary to that, in the body region, when viewed from the bottom, three layers are confirmed in the order of negative, positive, and positive. As mentioned above, the positive value near the upper boundary of the current appears as the result of the shear flow caused by the interaction with the ambient fluid. Besides, a pair of positive and negative values is estimated to be derived from distribution of the horizontal velocity u. That is, the second term in definition formula of vorticity ∆u / ∆y is dominant, because the vertical velocity v is almostzero in a fully developed region like the body region. As evidenced by the comparison of the vorticity and horizontal velocity distributions, the height hm, where u takes maximum value, corresponds closely to the height where the vorticity value takes zero.
  ‌Fig. 4-9 Flow structures assuming constant front velocity Uf obtained from (a) optical timeline image and (b) spatio-temporal velocity distributions, and (c) shows the vorticity distribution‌
  Concentration profileAs discussed above, concentration profiles of the quartz particles can be obtained using echo information. The spatio-temporal echo distribution obtained from UVP is shown Fig. 4-10. As you cansee in the figure, the echo intensity is a specific output value depending on the UVP devices and it fluctuates dramatically around a certain value. The distribution was obtained by subtracting the overall average value from each echo intensity. The head region shows relatively larger fluctuations in comparison with the body region. The mean-square amplitude of echo signal <V2> is required to obtain the concentration profiles, the schematic image of the procedure is shown in Fig. 4-11.
  ‌Fig. 4-10 Spatio-temporal echo distribution obtained by UVP
  ‌Fig. 4-11 Schematic image of the procedure to obtain average amplitude values <V2>Since the pixel noise is contained in the echo distribution, in this study, the averaged amplitude values by means of 5  5 measurement points were used for the calculation of <V2>, and the expression is as follows2 22 1  V i, j‌25    V i  i, j  j   . ( 4-14 )i2 j 2where i and j represent the coordinates in the time direction and the spatial direction of the measurement points, respectively. Using the value of <V2>, the spatio-temporal distribution of bulk density H calculated from the volume fraction from Eq.( 4-14 ) is shown in Fig. 4-12. As you can see, this method is under development and it is not yet possible to output high resolution and high accuracy distributions. Even now, many researches have been done to determine each variable as mentioned inthe previous section (e.g. Guerrero and Vittorio, 2018; Povey, 1997). However, the obtained distribution also gives us helpful information for understanding the flow structures of the turbidity current. In the density distribution, it is confirmed that the head region and bottom region near the bed contains high density corresponding to almost 1008 kg/m3 of initial density. Because the density of the quartz particles is 2650 kg/m3, they are heavier than water. Therefore, it is a reasonable result that the density of the particle-suspended layer becomes larger as the height y is lower. In addition to that, as estimated from the vector field, in the head region, a bottom layer is raised from near the bed to near the upper boundary by the rolling-up structure.
  ‌Fig. 4-12 Spato-temporal density distribution calculated from echo valuesThe time-averaged volume fraction  and the root mean square value of the fluctuation component ´RMS in the body region of 35 < t < 70 s are shown in Fig. 4-13. Since this analysis does not contain an algorithm that returns a value of 0 or 1 at each measurement point, there is no influence of the Heaviside formula,
   'RMS, ( 4-15 )Paying attention to the  profiles, the value approaches the maximum value of 0.5% near the bottom. Also, the increase in the values is observed especially below the height of hm, where u takes the maximum value. In the profiles of fluctuation component ´RMS, that takes local minimum value at y ≈ 20 mm. At that height, the  profiles have an inflection point. In addition, at around y = 30 mm, the ´RMS profiles show a positive peak like the shape observed in u´RMS profiles shown in Fig. 4-6. From these results, it is expected that the fluctuation of the volume fraction is closely related to the fluctuation of the horizontal velocity component. That is to say, it seems that the specific flow structure of the turbidity current as a turbulent flow contributes to such profiles of the volume fraction.
  ‌Fig. 4-13 Time-averaged volume fraction in the body region of 35 < t < 70 s‌
  Momentum equation in the body regionIn this section, the body region of the current, which can be treated as a quasi-steady flow, is discussed using a momentum conservation equation. For two–phase flows, two–fluid model has been used to reveal flow structures in wide fields. Especially in the case of dispersed bubbly flow, a momentum conservation equation based on two–fluid model has been used (e.g. Murai and Matsumoto, 2000; Kitagawa et al., 2001). The momentum conservation equations which can be applied to this study are explained below and we aim to quantify the flow structure in the body region.The momentum conservation equation for a two–phase flow by the two–fluid model is described asfL LuL   f u u  f p  M‌ M , ( 4-16 )t L L L L L LL SL‌fS SuS   f  u u   f p  M  M , ( 4-17 )t S S S S S SS LSwhere suffix L and S denote liquid and solid phase. The local time-dependent variables f, u, p are volume fraction, velocity vector, and pressure, respectively. The terms MLL and MSS are the inner phase momentum transport, which is modeled in the case of particle-laden flow as below:MLL  MSS   uL  uLT  2  u3 L  fL Lg‌( 4-18 )   u  u T  2  u   f  g S S 3 S  S S where  and g are viscosity and the acceleration vector of gravity, respectively. Here, the velocity vectors uL and uS can be treated as the same value u according to Stokes number which is a dimensionless number characterizing the behavior of particles suspended in a fluid flow: d 2USt p p 018W L0( 4-19 )where U0 and L0 are representative flow velocity and length, and their values are substituted by the time-averaged values of the layer-averaged velocity U and zero-velocity height ht shown in Table 4-1. As the result of calculation, the value of St is in order of 10-5 (St << 1), which means that the motion of the suspended particles is traceable to the fluid flow in this experimental case. Therefore, the Eq.( 4-18 ) is rewritten asΜLL  MSS    u  uT  2   u  f  g  f  g ( 4-20 )H  3  L L S S where  is the effective viscosity of particle-suspended fluid. In case of clay-free dispersions of sand, the value of effective viscosity was given by Davidson et al. (1977) asH  W 11.35 2.5 , ( 4-21 )where  is the volume fraction of the quartz in this case. Then, the terms MGL and MLG are mutual phase momentum transport. These two terms have the following relationship according to Newton’s third law:MSL  MLS  0.( 4-22 )Taking sum of Eq.( 4-16 ) and Eq.( 4-17 ), the equation of total momentum conservation is described by fL L  fS S u   f   f  uu  ft f p   L L S Su  uT  2  u   f  f . ( 4-23 )gL S H 3  L L S S The fluid is composed of just the quartz particles and water, so there is s relationship asfL  fS  1, ( 4-24 )fL L  fSS  H ; ( 4-25 )hence,Hu    uu  p    u  uT  2  u   g . ( 4-26 )t H H 3  H Treating the quartz-suspended fluid as incompressible fluid yieldsHu    uu  p  2  u   g . ( 4-27 )t H H HAs seen in the previous section, the horizontal velocity component is dominant in the body region of the turbidity current. Focusing on streamwise direction (horizontal velocity), the momentum conservation equation expressed by componential velocity is described by u  uu  uv  uw p  2 u 2 u 2 u ‌ H  H  H  H     H  H  H   Fg , ( 4-28 )t x y zx x2y2z2 where Fg denotes the external force caused by the gravity. The following variable separation into time- average and fluctuation components is carried.H  H  'Hu  u  u' v  v  v' w  w  w'( 4-29 )    'H H Hwhere u, v, and w stand the streamwise velocity, vertical velocity, and spanwise velocity, respectively. By substituting these equations into Eq.( 4-28 ) and averaging overall time for the body region 35 < t < 70 s, the following equation is obtained:  uu   u'u'  2' u' u  ' u'u' x H H H H   u v   u'v'  ' v' u  ' u' v  ' u'v' y H H H H H  u w  u' w'  ' w' u  ' u' w  ' u' w' . ( 4-30 )Hz H H H H HH2 u  ' u' Hx2 2 u  ' u' Hy2 2 u  ' u' HHz2 p  Fx gIn a well-developed flow, there is no mean vertical velocity ideally and no mean spanwise velocity. The gradient of momentum in the streamwise and spanwise directions is also negligible;therefore,y 2  u'v'  ' v' u  ' u'v'  HH H Hu  ' u' Hy2p x  Fg .( 4-31 )This equation is rewritten by p    H u' u' ‌H   u'v'  ' v' u  ' u'v'   F . ( 4-32 )x y  yy H H H  g Eventually the shear stress in the case of turbidity current consists of five terms as follows:  H u , ( 4-33 )1 y u' ' 2  H , ( 4-34 )y3 H   u'v' , ( 4-35 )4 H  ' v' u , ( 4-36 )5 H  ' u'v' . ( 4-37 )In the study by Middleton (1966), it is reported that the shape and velocity in the head region areadequately independent of the slope of the flume in case with angle < 2.3°. Hence, the term of Fg was assumed to be negligible, because the experimental flume was inclined at just 1.38 ゚ in the presentstudy.The computed results of the five components of shear stress are shown in Fig. 4-14, where the shear stresses from the effect of viscosity 1 and 2 are shown in Fig. 4-14(a) and the three components of turbulent shear stress 3, 4, and 5 are shown in Fig. 4-14(b). These values are computed from the experimental results obtained by a pair of UVPs in the body region of the current, where the averaging window is 35 < t < 70 s. The shear stresses shown on the horizontal axis are normalized by wall shear stress W estimated from the logarithmic fitting for the experimental results of velocity distribution according to the law of the wall as discussed above, and the y axis denotes normalized height by the zero-velocity height ht. As you can see, the shear stress of 1 shows the maximum value of 1/ W ≈ 0.3 near the bed and the value is largest in the whole distributions and thevalues of 2 is much smaller than 1. In the profiles of the three components of turbulent shear stress,Reynolds shear stress 3 is dominant, and the distributions of 4 and 5 have small fluctuation of thevalues, but the values are much smaller than 3. In the previous study about gas–liquid two–phase cannel flow, it was reported that the profiles of 4 take negative values near the top, which was caused by the fluctuation correlation between the local void fraction and the vertical liquid velocity (Murai et al., 2006). Considering that result, it is expected that the profiles of 4 in this study show different distribution from that in case of single–phase flows. Actually, however, the noticeable difference cannot be observed. The reason for deriving such a result seems to be that the local volume fraction of suspended particle ( ≤ 0.5%) is much smaller than that in the case of bubbly flow with around 60% in the maximum void fraction.
  ‌Fig. 4-14 Computed profiles of the shear stress; (a) shear stresses caused by viscosity and (b) turbulent shear stressesIn addition, the distribution of viscous stress 1 and Reynolds stress 3 show characteristic shapes. Ideal distributions of 1 and 3 in a channel flow are shown in Fig. 4-15 (Kida and Yanase, 1999 (in Japanese)), and such an axisymmetric flow contains point-symmetric distributions of shear stresses.Contrary to that, the distributions of 1 and 3 in this study do not have such profiles. The difference about 1 between the present study and a channel flow is the upper boundary, where it is an ambient fluid in this study and a solid wall in the channel flow. Therefore, the result of 1 in this experiment, which the upper part takes negative values overall, is due to the shear flow by the ambient fluid. On the other hand, the Reynolds shear stress 3 takes a value of close to zero near the bed (inner region: y < hm). The presence of particles may suppress the fluctuation of velocity components as described above.
  ‌Fig. 4-15 Schematic of shear stress in case of single-phase channel flow (Kida and Yanase, 1999(in Japanese))The total shear stress, which is the sum of five shear stress components (1, 2, 3, 4, and5), is shown in Fig. 4-16. The computed result taking the value close to zero in the range of0.3 < y/ht < 0.8 indicate that the values of 1 and 3 are nearly canceled out. The total shear stress at y = 0.561 mm, the lowest measurement point by UVP in this study, is 0.030 Pa. That value is the same order as 0.086 Pa of the wall shear stress W estimated from the fitting result for the velocity distribution according to law of the wall.Additionally, the wall shear stress is roughly estimated just from the geometric configuration in case of this experimental facility. An extremely simplified experimental flume is shown in Fig. 4-17 to discuss the force balance in a hydrostatic fluid pressure. In the figure, A and B denote area of the bottom and the cross section of the flume, respectively. Then, the wall shear stress working at the bed is described as W and the pressure difference p is generated by the density difference of the both fluids between the left-hand side and the right-hand side. Considering the force balance of these physical quantities, the following equation is obtained,‌ W  A  p  B . ( 4-38 )Here, it can be assumed that the pressure difference p is generated by the density difference between the particle-suspended fluid and ambient fluid, and their density are shown as H and W, respectively. Using such an assumption, the following relationship is obtained.p  g B  2H  WB g . 2‌( 4-39 )As mentioned above, the test section is a rectangle flume with 4,548 mm in total length, 210 mm in height and 143 mm in width, so the values of A and B can be calculated easily. Substituting Eq.( 4-39 ) in Eq.( 4-38 ), the following equation is obtained,B2 W  2A g . ( 4-40 )The calculated value of the wall shear stress is 0.055 Pa, assuming that other forces like friction on the side walls are not working. This value shows the same order as the shear stresses calculated from the velocity distribution discussed above (W = 0.086 Pa according to law of the wall). From this extremely simplified validation result, it is confirmed that the wall shear stress estimated form velocity distribution shows the reasonable order of value.
  ‌Fig. 4-16 Sum of five shear stress components
  ‌Fig. 4-17 Schematic image of simplified experimental flumeAccording to Eq.( 4-32 ), vertical profiles of the pressure gradient acting on the body region of the current (35 < t < 70 s) can be calculated by differentiating the total shear stress with respect to the height y direction, and the obtained distribution is shown in Fig. 4-18. As mentioned above, since the shear stresses are balanced inside the flow (outer region: hm < y < ht), the pressure gradient shows large values only near the upper (y ≈ ht) and lower boundaries (inner region: y < hm).Such a phenomenon seen in the region 0.3 < y / ht < 0.8 (outer region) shows a negative momentum transfer against the mean velocity gradient, which is working to keep a stable stratification. The negative momentum transfers are reported in the previous experimental studies on turbulent transport phenomenon in density stratification, for instance, saline density stratified flows and wind tunnel experiments (e.g. Komori et al., 1983; Piccirillo and van Atta, 1997). This study reveals the existence of such the phenomenon in the turbidity currents containing fine particles, that is, solid– liquid two–phase turbulent flows. In this case, as the result of the turbulent energy balanced with the potential energy of the suspended particles, it is expected that the particles are being mixed by the velocity fluctuation components. It has the effect to suppress the settling of the suspended particles and to keep H at a roughly constant value. Additionally, in the inner region of the body region, theparticle-laden flow with high density H is extruded from the rear by a pressure gradient shown inFig. 4-18, that contributes to a large amount of particle transporting action.
  ‌Fig. 4-18 Computed pressure gradient
3. ‌Comparison of flow structures between quartz-suspended currents, and opalin-suspended currents, and two–layer turbidity currentsIn this section, the differences of flow characteristics between one–layer turbidity currents including quartz-suspended currents and opalin-suspended currents and two–layer turbidity currents are discussed. The behavior of currents with more clearly different initial conditions from each other than the conditions discussed above is described. The currents of total nine experimental cases are mainly discussed in this section. These initial conditions and front velocity calculated by the pattern matching method are summarized in Table 4-5. In Case 19, the quartz-suspended fluid contained in tank A was colored by a red ink to distinguish from the fluid contained in the tank B.‌Table 4-5 Initial conditions and calculated front velocity
  ‌
  Experimental snapshots of the currentsOutline motions of the two–layer turbidity currents invading into the ambient fluid are monitored as sequential images from t = t0 to t0 + 10.5 s. The images were taken by the digital camera installed in front of the flume for Case 19, 20, and 22 (Fig. 4-19, Fig. 4-20, and Fig. 4-21). In the images of Case 19, the lower layer is composed of water colored by red ink and quartz particles with 2.0% in volume fraction, but it is difficult to distinguish the red lower layer from white upper layer. On the other hand, the sequential images of experimental Case 20 and 22 show the rough boundary distinguishing the upper layer and lower layer due to the color differences of those two layers, which means that the quartz-suspended layer has white color and the opalin-suspended layer has black color. From these images, it is confirmed that this experimental facility is able to generate two–layer turbidity currents. Comparing the sequential images of the three cases, flow structures of them can be discussed qualitatively. The large vortex structures attributed to the K-H instability can be observed near the upper boundary in the all cases. Although the quantitative discussion cannot be done, the wavenumber of the interfaces between the particle-laden fluid and the ambient fluid formed by some visible large vortices in the images almost does not vary under the three conditions. In addition, the rolling-up structures in the head region is particularly well visible in the images of the Cases 20 and 22.
  ‌Fig. 4-19 Snapshots of quartz-suspended turbidity currents of Case 19 from t = t0 to t0 + 10.5 s
  ‌Fig. 4-20 Snapshots of quartz-suspended turbidity currents of Case 20 from t = t0 to t0 + 10.5 s
  ‌Fig. 4-21 Snapshots of quartz-suspended turbidity currents of Case 22 from t = t0 to t0 + 10.5 s‌
  Inner flow structuresSpatio-temporal distributions of horizontal and vertical velocity components (u and v) obtained by UVP in each case are shown in Fig. 4-22 and Fig. 4-23, where (a) left, (b) middle, and (c) right panels represents the results of quartz-laden currents, opalin-laden currents, and two–layer currents, respectively. The axes denote the height (y) from the bed and the elapsed time (t) from the release ofthe gate 1. As the color scale of velocity magnitude, u is dominant in the flow in those cases too. Paying attention to the cases of (a) quartz-laden currents and (b) opalin-laden currents, the initial density of the particle-suspended fluids becomes larger from the upper panels to lower, namely the density of the upper panels, middle panels, and lower panels are 1008, 1016, and 1032 kg/m³, respectively. Under each condition, it is confirmed that the horizontal velocity component u is increasing as the initial density of the particle-suspended fluids becomes larger form color contour in the panels. In addition, comparing the quartz-laden current and the opalin-laden current having the same initial density, it also can be confirmed that the quartz-laden current exhibits larger velocity values as a rough tendency. Contrary to that, the bulk density of (c) two-layer turbidity currents is constant value within experimental Case of 19, 20, and 22. The calculated bulk density considering the volume difference between the tank A and the tank B is 1023 kg/m³. In those three cases, the spatio-temporal distributions of horizontal velocity u show similar values and trends between the experimental Cases 19 and 20, but the distribution of Case 22, where the bottom is composed of the opalin layer, shows lower velocity values in comparison with those two cases (Case 19 and Case 20). From these comparisons, it is predicted that the opalin-laden currents and two-layer currents, where the bottom layer is the opalin, have a tendency to have lower velocity values comparing with the quartz-laden currents. The dot lines shown in Fig. 4-22 indicate the expected arrival time t* defined in the previous section at xI = 1005 mm that is the cross point of two ultrasonic measurement lines. Those expected arrival time summarized in Table 4-6 roughly coincides with the results measured by UVP in the cases discussed in this section.
  ‌Fig. 4-22 Spatio-temporal distributions of horizontal velocity component: (a) quartz-laden currents, (b) opalin-laden currents, and (c) two–layer turbidity currents
  ‌Fig. 4-23 Spatio-temporal distributions of vertical velocity component: (a) quartz-laden currents, (b) opalin-laden currents, and (c) two–layer turbidity currentsIn order to estimate the wall shear stress working in the body region, the body region should be separated from the head region. In these cases, however, the front velocity Uf varies with the initial conditions, and thus it is difficult to define the separation point between the head and the body of the currents from elapsed time t. Therefore, the length of the head is estimated from these experimental results. Concretely, combining the front velocity value Uf and the duration time evaluated from velocity distributions, the length of the head region can be estimated roughly. The duration time means passing time of the head of the currents under the ultrasonic measurement lines. Uf calculated by pattern matching method is used for the analysis. For qualitative understandings, the flow character defining the separation point between the head and the body is defined as the wake behind the head in a previous research. For example, such a flow character can be observed in the section of 4.2. As the results of such a process, the length of the head region in this study is estimated to be about 500 mm. Furthermore, based on empirical judgment using experimental results, in the experimental conditions and the experimental facility of this study, it was estimated that the body region, which is the wake of the head region, propagates over at least a length corresponding to 2000 mm. Using these estimated lengths, each time for the averaging window is defined as,t1  t*x xI 500 mmt2  t*x xI 2500 mm xI  500 ,U f xI  2500 .U f( 4-41 )( 4-42 )The time-averaging widow is from t = t1 to t2, so the window sizes are different in each experimental case as shown in Table 4-6.‌Table 4-6 Summary of the estimated arrival times
  The profiles of the time-averaged u of each case are shown in Fig. 4-24, with the averaging window is from t = t1 to t2 calculated for each experimental case. In the panels of Fig. 4-24(a) and (b), the results in the cases of the quartz-laden currents and opalin-laden currents are shown and it isconfirmed that the values of the horizontal velocity components u also increase according to the initial density of the particle-suspended fluids (in these panels, the blue, green, and red symbols denote the experimental results of particle-laden flows with H = 1008, 1016, and 1032 kg/m3, respectively.).
  ‌Fig. 4-24 Time-averaged profiles of horizontal velocity component in the body region (t1 < t < t2) of (a) quartz-laden currents, (b) opalin-laden currents, and (c) two–layer currentsThe representative horizontal velocity u and the heights of the currents calculating from the time- averaged velocity profiles are summarized in Table 4-7. Here, umean denotes the mean velocity based on the integration from y = 0 to ht. The dimensionless numbers to evaluate the general flow structures like Reynolds number (the ratio of inertial forces to viscous forces), densiometric Froude number (the ratio of inertia to the gravity), and Richardson number (the ratio of buoyancy term to inertia) are defined in the study by Kneller et al. (1999) asumean  1 ht u( y)dy , ( 4-43 )ht y 0Re  Humean D , ( 4-44 )Hg'DFr'  umean , ( 4-45 )Ri g'Du2mean, ( 4-46 )where, D is the current thickness and the quantitative clear definition of D is not expressed in the previous studies. In this study, therefore, ht was used as the current thickness instead of D. Calculated those dimensionless numbers in the body region are shown in Table 4-8. The Reynolds numbers in the all cases have values of (103), so the currents are treated as turbulent flows. Comparing the one– layer currents between the quartz-laden and opalin-laden currents having same initial density, the Reynolds number is larger in cases of quartz-laden currents, which means that the viscosity moreaffects the flow behavior in cases of opalin-laden currents. Based on the definition of Fr´ and Ri, they have a relationship derived asRi 1Fr' 2. ( 4-47 )Paying attention to the values of Fr´ and Ri, the quartz-laden currents and two–layer currents in Case 19 and Case 20, where the bottom layer is the quartz-suspended fluid, show higher Fr´ values exceeding 0.7 than the remaining cases. Therefore, the Ri values show the opposite trend. It means that, in opalin-laden currents and two-layer currents of Case 22, where the bottom layer is the opalin- suspended fluid, buoyancy term is dominant.
  ‌Table 4-7 Representative horizontal velocity and heights on the currents‌Table 4-8 Summary of dimensionless numbers; Reynolds number, densiometric Froude number, and Richardson number
  In order to estimate the wall shear stress in the body region y < hm of these currents, the logarithmic fitting based on law of the wall, which was already discussed in chapter 4.1.3 was applied to these profiles. The fitting results are shown in Fig. 4-24, here the von Karman constant  is substituted 0.41 assuming a smooth bed. In the panels shown as single logarithmic charts, each symbol indicates the experimental values and each solid line indicates the fitting results, and the distributions of both values show generally good agreement. The fitting results obtained by this method are summarized in Table 4-9. As the value of the maximum velocity of the horizontal component u increases, the value of u* increases. Therefore, the value of u* is normalized by umax and those values of u*/ umax take an averaged value of 0.114 in these nine cases not depending on the initial conditions. Utilizing this property, the friction coefficient Cf was calculated using the following equation u * 2Cf  2umax . ( 4-48 )The calculated values of Cf are shown in Table 4-9. Because of u*/ umax ≈ 0.114, the values of Cf take around 0.0260, which was found to be one order larger than the values measured by the bubbly flows in case of horizontal channel flows (Murai et al., 2006; Murai et al., 2007) and towing tank experiment (Park et al., 2018).
  Laboratory for Flow Control, Hokkaido University
  Fig. 4-25 Fitting results for time-averaged profiles of horizontal velocity component in the body region of (a) quartz-laden currents, (b) opalin- laden currents, and (c) two–layer currents‌Table 4-9 Fitting results for experimental velocity values in the body region
  ‌
  Interfaces in two–layer turbidity currentsIn the two–layer turbidity currents of the experimental Case 19, 20 and 22, the detection of two interfaces on ambient fluid–upper layer and upper layer–lower layer was tried to evaluate the flow behaviors. Optical timeline images for each case are shown in Fig. 4-26, which shows the time variation at xI = 1005 mm, corresponding to the cross point of two ultrasonic measurement lines. Since these images are taken from the front of the flume, the visible region is the flow state near the front side wall. In spite of that, the interfaces can be observed in these cases from the color differences between two layers. Especially, the interface between the upper layer and lower layer is easy to be detected in Case 20. In the image of Case 20, it is confirmed that the lower layer composed of quartz- suspended fluid reaches at y ≈ 60 mm in the head region, then the interface is distributed around y = 20 mm in the body region, gradually decreasing from about t = 30 s. After t = 40 s, the flow returning from the end of the flume can be observed. Here, it is confirmed that the shape of the current head is maintained and that is invading under the current flowing from the left-hand side as shown like Fig. 4-27. From this behavior, it is predicted that the density of the head region is larger than that of the body region flowing from the left-hand side.To compare the arrangements of the two layers (the lower layer and the upper layer) with these three Cases 19, 20 and 22, timeline image at x = xI of the case 24 is shown in Fig. 4-28. In the Case 24, the current is composed of the lower quartz-suspended layer with H = 1032 kg/m3 and theupper opalin-suspended layer with H = 1008 kg/m3. The bulk density considering the volumedifference between the tank A and the tank B is 1019 kg/m3. In that case, the density difference between the lower layer and the upper layer is larger than the three cases mentioned above (Cases 19, 20 and 22), so the lower layer propagates much faster than the upper layer in Case 24. In the Fig. 4-28,the upper layer is not clearly confirmed in the head region, and the black upper layer composed of opalin-suspended fluid can be seen at around t ≥ 20 s.
  ‌Fig. 4-26 Timeline image at the cross-point of two ultrasonic beams, Case 19 is composed of quarts layers, Case 20 is composed of lower quartz layer and upper opalin layer, Case 20 is composed of lower opalin layer and upper quartz layer
  ‌Fig. 4-27 Schematic drawing of propagation of the head region under the down stream flow
  ‌Fig. 4-28 Timeline image of Case 24 at the cross-point of two ultrasonic beamsIn order to evaluate the height of the interfaces, spatio-temporal echo distributions obtained by UVP were used. As mentioned above, in the simultaneous measurement of two UVPs, the ultrasonic interference noise occurs. To reduce the noise, the following series of processing (Fig. 4-29) was performed using the two echo distributions obtained from TDX 1 and TDX 2. In this processing, firstly, time correction method is applied to the spatio-temporal echo distribution too using front velocity Uf according to Eq.( 3-3 ) and Eq.( 3-4 ), since the positions of two ultrasonic transducers have a certain displacement outside the intersection point. As seen in Fig. 4-29(2), the echo distribution contains the ultrasonic interference noises distributed over time and space, which shows clearly large fluctuations compared with surrounding values. Next, therefore, the echo amplitude value which is the absolute value of the difference from the average value of the echo intensity, was calculated. Then, the smaller values are chosen for the reasonable values at each measurement point comparing two echo amplitudevalues obtained from TDX 1 and TDX 2, because the noise has an obviously large value of echo amplitude. As the result of this processing, the noise-reduced distribution of echo amplitude can be obtained like Fig. 4-29(5).
  ‌Fig. 4-29 A series of noise reduction method in echo distributionsThe echo amplitude distributions of Case 19, 20, and 22 obtained using the noise reduction method are shown in Fig. 4-30. The noises are still remaining in these panels, but it is greatly improved in comparison with raw data of echo intensity distribution obtained by UVP. Comparing the results of the three cases, there are interesting differences depending on the combination of the kind of suspended particles contained in two layers; the upper layer and the lower layer. In Case 19, both layers are composed of quartz-suspended fluid, and echo amplitude distribution show approximately uniform values in the current. Contrary to that, the distributions of Case 20 and 22 have two regions in the currents; one of them shows a relatively larger values and the other is not so. The region showing large values is the layer composed of opalin-suspended fluid. As mentioned above, in case where the particle diameter d is much smaller than ultrasonic wave length  (d/  << 1), Rayleigh scattering occurs. The scattering intensity increases as the particle diameter becomes larger in the cases of Rayleigh scattering. Since the center diameter of quartz and opalin particles are 12.2 and 18.9 m, respectively, it is reasonable that the oplin layer shows larger echo amplitude values. Using the characteristic, thedetection of two interfaces was examined for the experimental Cases 20 and 22.
  ‌Fig. 4-30 Spatio-temporal distributions of echo amplitude, Case 19 is composed of quarts layers, Case 20 is composed of lower quartz layer and upper opalin layer, Case 20 is composed of lower opalin layer and upper quartz layerTo measure the positions of interfaces quantitatively, the vertical profile of echo amplitude in Case 20 at t = 30 s is shown in Fig. 4-31 as a sample. As shown in the figure, the echo amplitude has large fluctuations, that makes it difficult to detect interfaces using these instantaneous profiles at each measurement time. In order to overcome this difficulty, polynomial fitting was applied to the echo amplitude distributions at each measurement time, although an arbitrary property remains in this method. In this study, tenth-order polynomial function was examined for the fitting and the results are also shown in Fig. 4-31 by a blue solid line. The fitting result takes maximum value at y ≈ 40 mm and the height corresponds to approximately the center position of opalin layer in comparison with the distribution shown in Fig. 4-30. This method was also applied to the echo distribution on Case 22 and the results are shown in Fig. 4-32. In this result for the Case 22 shown by blue line too, the fittingcurve takes maximum value at y ≈ 20 mm corresponding to roughly the center of opalin layer. Therefore, it is expected that the values obtained from the polynomial fitting provide beneficial information to detect interfaces. The polynomial fitting was applied to the whole measurement time of echo amplitude in the Case 20 and 22, and the results are shown in Fig. 4-33.
  ‌Fig. 4-31 Vertical profile of echo amplitude in Case 20 at t = 30 s
  ‌Fig. 4-32 Fitting results for the Case 20 and 22Using the fitting results, two kinds of threshold values T1 and T2 for the echo amplitude values E(y, t) were defined to determine two interfaces. The threshold value T1 is used for the interface between the upper layer and the lower layer, using the slope of the echo amplitude for the direction of height. In the two cases, the height of the opalin layer changes depending on the experimental case, so the values of T1 have different definition for each case likeT  Max. dE  y,t   (for Case 20) , ( 4-49 )1  dy T  Min. dE  y,t   (for Case 22) . ( 4-50 )1  dy The height y1 satisfying the above conditional expression at each measurement time is treated as the interface 1. In addition, to determine the interface 2 between the upper layer and the ambient fluid, thevalues of T2 was defined as followsT  E  y1,t   Min.E t  , ( 4-51 )2 2where Min.[E(t)] means the minimum value of the echo amplitude at t, and its value becomes nearly zero at most time t. The echo amplitude distribution was examined from the top at every measurement time, and the height y2 at which the echo amplitude exceeded the threshold at the beginning was defined as the interface 2.
  ‌Fig. 4-33 Spatio-temporal distributions of polynomial fitting results for echo amplitude, Case 20 is composed of lower quartz layer and upper opalin layer, Case 22 is composed of lower opalin layer and upper quartz layerUsing those threshold values T1 and T2 for the echo amplitude values, the two interfaces at each measurement time were detected and the results are shown in Fig. 4-34. The detected interfaces are shown on the timeline image at x = xI. Comparing the detected interfaces with optical images, the interface 2 corresponding to the upper interface is in good agreement in both cases of 20 and 22. Onthe other hand, there is room for discussion about the interface 1 corresponding to the lower interface. In particular, at around 15 < t < 25 mm in the Case 22, a discontinuous distribution about the height of the interface 1 determined from the echo amplitude is observed. Here, about the shape of the echo amplitude distributions, the existence distribution of opalin having a large diameter is dominant. That is, according to the above definition of T1, it is possible to indicate the lower limit of the height at which the opalin exists in the upper layer for the Case 20 and the upper limit of the height at which the opalin exists in the lower layer for the Case 22. Considering this characteristic, the distribution at around 15 < t < 25 mm in the Case 22 shows that the rolling-up head region of the currents is composed of not only the upper layer but also the lower layer.Additionally, these results can support qualitatively the rough validation of the density distribution shown in Fig. 4-12 estimated from the echo amplitude obtained by UVP. In that density distribution, the head region and the lower body region has relatively large values of the density, and near the upper boundary region at about 25 < t < 40 s also shows a large value sparsely. Such high- density values near the upper boundary seem related to the rolling-up structure in the head region. Namely, the distribution is attributed to the raised high-density fluid which is transported from near the bed to the upper boundary region. That hypothesis can explain the tendency of the density distribution shown in Fig. 4-12.
  ‌Fig. 4-34 Distributions of the detected two interfaces using echo amplitude values, Case 20 is composed of lower quartz layer and upper opalin layer, Case 22 is composed of lower opalin layer and upper quartz layer‌
  Summary of flow behaviors in the some kinds of turbidity currentsThis part summarizes the comparison of the flow behaviors between the quarts-laden currents, opalin- laden currents, and two-layer currents.In the discussion about the front velocity Uf, it is revealed that the quartz-laden currents propagate faster than the opalin-laden currents with the same initial density H as a particle-suspended fluid. In addition, in cases of two–layer currents, the bottom layer is dominant to determine the Uf. In all conditions, the friction coefficient Cf shows values of around 0.0260 in the case that the maximum horizontal velocity umax in the body region is used for normalization, because the relation of u* / umax ≈ 0.114 is confirmed in these experimental cases.Using echo amplitude distributions, it becomes possible to detect two interfaces, because the scattering intensity in the cases of Rayleigh scattering reflected by the fine particles varies depending on the diameter of the particles. In this study, therefore, the echo amplitude values based on the scattering by the opalin particles are larger than that by the quartz particles. Considering the results obtained from the interfaces distribution in Case 22, the rolled-up structure can be observed in the head region and it is revealed that the head is composed of not only upper layer but also lower layer as shown in Fig. 4-35(a). This tendency can be mentioned in the study of Gladstone et al. (2004), although their study focused on the saline density currents. As shown in Fig. 1-2, the lower layer and the upper layer mix in case of the dimensionless density ratio between the layers * ≈ 0.5 like theseexperimental cases. On the other hand, the dimensionless density ratio of the Case 24 is * ≈ 0.25,which means that the density difference between the two layers is larger than in the Case 22. In such a case, the lower layer propagates much faster than the lower layer, so the mixing of the two layers cannot be confirmed in the head region as shown in Fig. 4-35(b).At the bottom, the lower layer plays a major role in determining the friction occurring between the lower layer and the bed. Furthermore, near the head region of currents, there is an interaction between the current and the ambient fluid, water. As a result of the rolled-up structure like shown in Fig. 4-35, only the lower layer is related to the most of these interactions. Therefore, in a discussion of front velocity and friction acting on the bed, it is determined by the kind of the lower layer. This hypothesis well describes the obtained experimental results. Under these experimental conditions, it is considered that the front velocity is not so influenced by the interaction based on the K-H instability seen near the upper boundary.
  ‌Fig. 4-35 Schematic drawing of the flow behavior in case the two-layer turbidity currents, (a) in case of * ≈ 0.5 and (b) in case of * ≈ 0.25
‌Conclusion
This study was carried out for elucidating the long-distance propagation mechanism of turbidity currents, which are sediment-gravity flows and driven by the density difference between particle- suspended fluid and ambient fluid. Such turbidity currents have complicated inner flow structures and they cannot be measured by optical approaches because of their opacity. In order to evaluate their flow behaviors quantitatively, it was required to expand the utility of ultrasonic measurement techniques to be applied to solid–liquid two–phase flows. The conclusion about results and knowledge obtained in this study are summarized below.Ultrasonic measurement technique
- Ultrasonic velocity profiler (UVP) was applied to the measurements for particle-laden flows. The diameter of suspended particles d was order of 10 m and the volume faction of particles  was in range of 0.01 <  < 15%. This condition is difficult for UVP measurement to work well, because the size of d is too small to work as a reflector of ultrasonic waves with 4 MHz in basic frequency in water and large volume fraction promotes attenuation of the ultrasonic waves. Nevertheless, in case of quartz particles with 12.2 m in center diameter, the reasonable velocity distributions could be measured by UVP in 0.1 <  < 5%.
- Use of a pair of UVPs makes it possible to obtain both velocity components of horizontal u and vertical v. In addition to that, this study suggested the time-correction method at each measurement position using representative velocity.
- When the particle diameter d is much smaller than wavelength of ultrasound  (d /  << 1),Rayleigh scattering, which is almost isotropic scattering, occurs. In such a case, the estimation of particle volume fraction was examined in this study. The novelty of this method is that echo obtained by UVP was applied to experimental turbidity current and boundary condition was modified for turbidity currents. Although highly accurate measurement results have not yet been obtained, it seems to become of great help for the elucidation of the flow structures of solid-liquid two-phase flow.
- In the condition of d /  << 1, scattering intensity increases as the diameter of suspended particlesbecomes larger. Using this property, in cases of two-layer turbidity currents, it was successful to detect two interfaces of ambient fluid-upper layer and upper layer-lower layer.Flow structures of gravity currents
- A quantitative method to track the head positions of the gravity currents using the pattern matching method for consecutive experimental images was established, and the front velocity Uf, which is the advection velocity of the current, was calculated for each experiment. As the results,it was found that the turbidity currents containing quartz particles propagates 19.0% faster than the saline density currents, for the same initial density. Additionally, the turbidity currents containing quartz particles propagates 38.5% faster than turbidity currents containing opalin particles. From this, it was confirmed that the front velocity Uf varies greatly due to the interaction caused by the suspended particles.
- The flow structure of the turbidity currents containing quartz particles and the saline density currents were compared. Comparing the height of the two gravity currents in the body region, it was found that the height of the turbidity current was lower. This is because the diffusion is predominant in the density current, whereas, in case of the turbidity current, settling of the particles is also included. As the results, the distribution of the velocity fluctuation component greatly changed. In case of the turbidity current, large fluctuation occurred at y > hm, but in case of the density current, the tendency was reversed. From this, it is thought that the existence of the particles suppresses the velocity fluctuations in the vicinity of the bed. Besides, the particles are mixed by the turbulent velocity fluctuations in the region above that (y < hm). That mixing acts as an effect giving potential energy to particles, namely the supplemental driving force is supplied for the current, because the density difference is kept. That is why, despite the fact that the friction coefficient Cf is smaller in case of the density current, the turbidity current can propagate faster the density current.
- Flow behavior of the turbidity current containing quartz particles as the solid–liquid two–phase flow was analyzed, using momentum conservation equation in the body region. From thecalculated results of each shear stress distribution, it was found that viscous stressdHudyandReynolds shear stressH u' v' are dominant as in cases of single–phase flows. In cases of theHturbidity current, because the volume fraction  of quartz particles is small, the value of ' v' uwas found to be sufficiently smaller thanH u' v' . On the other hand, it was confirmed that sincethedHudyandH u' v' are canceled out at outside the bottom and the upper boundary, thepressure gradient hardly acts. Namely, a negative momentum transfer against the mean velocity gradient, which is working to keep a stable stratification, is confirmed in the outer region (y > hm) of the body region. It is expected that the particles are being mixed by the velocity fluctuation components, that suppresses the settling of the suspended particles and to keep H at a roughly constant value.
- Analysis of two-layer turbidity currents showed that the type of particles suspended in the lowerlayer determines the front velocity Uf, when the bulk density H as initial condition is equal. In the two–layer turbidity currents composed of different kinds of particle-suspended fluid, it was confirmed that the lower layer was rolled up in the head region and the upper layer was wrapped. Therefore, in these conditions, not only the friction with the bed but also the interaction with the ambient fluid in the vicinity of the head is dominated by the lower layer. From this, it was found that the front velocity Uf was determined by the friction at the bottom and the interaction at the head region, and the influence of shear flow at the upper boundary in the body region does not work much.
- In cases of the turbidity currents, it was confirmed that the relationship between the friction velocity u* and the maximum velocity umax in the body region was u*/umax ≈ 0.114 in every experimental case. It was also confirmed that the friction coefficient Cf ≈ 0.0260 was obtained using this relationship, the obtained values of Cf are one order larger than that measured in gas– liquid two–phase flows.
‌Bibliography

Altinakar, M. S., Graf, W. H., Hopfinger, E. J., "Flow structure in turbidity currents." Journal of Hydraulic Research 34.5 (1996): 713-718.

Azpiroz-Zabala, M., et al. "Newly recognized turbidity current structure can explain prolonged flushing of submarine canyons." Science advances 3.10 (2017): e1700200.

Barr, D. I. H. "Densimetric exchange flow in rectangular channels III. Large scale experiments." La Houille Blanche 22 (1967): 619-632.

Cesare, G. D., Anton, S., Felix, H., "Impact of turbidity currents on reservoir sedimentation." Journal of Hydraulic Engineering 127.1 (2001): 6-16.

Chamoun, S., Cesare, G. D., Schleiss, A. J., "Managing reservoir sedimentation by venting turbidity currents: A review." International Journal of Sediment Research 31.3 (2016): 195-204.

Davidson, J. F., Harrison, D., Carvalho, J. R. F. G. D., "On the liquid like behavior of fluidized beds."

Annual Review of Fluid Mechanics 9.1 (1977): 55-86.

Dong, X., et al. "Oil–water two-phase flow velocity measurement with continuous wave ultrasound Doppler." Chemical Engineering Science 135 (2015): 155-165.

Downing, A., Thorne, P. D., Vincent, C. E., "Backscattering from a suspension in the near field of a piston transducer." The Journal of the Acoustical Society of America97.3 (1995): 1614-1620.

Ellison, T. H., Turner, J. S., "Turbulent entrainment in stratified flows." Journal of Fluid Mechanics 6.3 (1959): 423-448.

Fisher, F. H., Simmons, V. P. "Sound absorption in sea water." The Journal of the Acoustical Society of America 62.3 (1977): 558-564.

Gladstone, C. et al., "An experimental investigation of density‐stratified inertial gravity currents." Sedimentology 51.4 (2004): 767-789.

Guerrero, M., Vittorio, D. F., "Suspended sediment assessment by combining sound attenuation and backscatter measurements–analytical method and experimental validation." Advances in Water Resources 113 (2018): 167-179.

Hitomi, J. et al. "Ultrasound Flow-Monitoring and Flow-Metering of Air-Oil-Water Three-Layer Pipe Flows." IEEE Access5 (2017): 15021-15029.

Hitomi, J., Nomura, S., Cesare, G. D., Takeda, Y., Park, H. J., Tasaka, Y., Murai, Y., "Flow Monitoring of Particle-laden Flows Combining Ultrasonic Doppler and Echo Intensity Profiling Techniques.", Proc. of the 11th International Symposium on Ultrasonic Doppler Methods for Fluid Mechanics and Fluid Engineering 11th ISUD, September 5-7, (2018), Berlin, Germany,

pp. 75-78

Jamshidnia, H., Takeda, Y., "An experimental study of the effect of a baffle on the flow structure in a rectangular open channel using UVP." Journal of Fluid Science and Technology 5.3 (2010):

542-557.

Kang, L., Feeney, A., Dixon, S., "Flow measurement based on two-dimensional flexural ultrasonic phased arrays." Proceedings of Meetings on Acoustics 6ICU. Vol. 32. No. 1. ASA, 2017.

Keulegan, G. H., "An experimental study of the motion of saline water from locks into fresh water channels." Nat. Bur. Stand. Rept. Technical Report 5168 (1957).

木田重雄, 柳瀬眞一郎. 乱流力学. 朝倉書店, 1999.

Kikura, H., Hamdani, A., Ihara, T., "Development of Two-dimensional Vector UVP with Phased Array." Development 28 (2016): 30.

Kitagawa, A., Murai, Y., Yamamoto, F., "Two-way coupling of Eulerian–Lagrangian model for dispersed multiphase flows using filtering functions." International journal of multiphase flow 27.12 (2001): 2129-2153.

北浦秀和, 村井祐一, and 武田靖. "UVP による渦輪の速度ベクトル場の計測 (流体工学, 流

体機械)." 日本機械学会論文集 B 編 76.772 (2010): 2143-2151.

Kneller, B. C., Bennett, S. J., & McCaffrey, W. D., "Velocity structure, turbulence and fluid stresses in experimental gravity currents. " Journal of Geophysical Research: Oceans, 104(C3) (1999), 5381-5391.

Kneller, B., & Buckee, C., "The structure and fluid mechanics of turbidity currents: a review of some recent studies and their geological implications. " Sedimentology, 47 (2000), 62-94.

Komori, S. et al. "Turbulence structure in stably stratified open-channel flow." Journal of Fluid Mechanics 130 (1983): 13-26.

Lee, T. H., & Hanes, D. M., "Explicit solution to the acoustic backscatter equation to measure the concentration of uniform, suspended particles. " Journal of Geophysical Research, 100(C2) (1995), 2649-2657.

Longo, S., et al. "Gravity currents produced by constant and time varying inflow in a circular cross- section channel: Experiments and theory." Advances in Water Resources 90 (2016): 10-23.

Malvern S. A., “Operational qualification for Mastersizer 3000 (MAZ3000) optical bench”, 2013.

Middleton, G. V. "Sediment deposition from turbidity currents." Annual review of earth and planetary sciences 21.1 (1993): 89-114.

Middleton, G. V. "Experiments on density and turbidity currents: I. Motion of the head." Canadian Journal of Earth Sciences 3.4 (1966): 523-546.

Murai, Y., Matsumoto, Y., "Numerical study of the three-dimensional structure of a bubble plume." Journal of fluids engineering 122.4 (2000): 754-760.

Murai, Y. et al. "Turbulent shear stress profiles in a bubbly channel flow assessed by particle tracking velocimetry." Experiments in Fluids 41.2 (2006): 343.

Murai, Y. et al. "Skin friction reduction by large air bubbles in a horizontal channel flow." International journal of multiphase flow 33.2 (2007): 147-163.

Murai, Y. et al. "Development of an ultrasonic void fraction profiler." Measurement Science and Technology 20.11 (2009): 114003.

Murai, Y. et al. "Ultrasonic detection of moving interfaces in gas–liquid two-phase flow." Flow Measurement and Instrumentation 21.3 (2010): 356-366.

Murakawa, H., Kikura, H., Aritomi, M., "Application of ultrasonic multi-wave method for two-phase bubbly and slug flows." Flow measurement and instrumentation19.3-4 (2008): 205-213.

Nogueira, H. I., Adduce, C., Alves, E., Franca, M. J., "Image analysis technique applied to lock- exchange gravity currents." Measurement Science and Technology 24.4 (2013): 047001.

Nomura, S. et al. "Sediment mass movement of a particle-laden turbidity current based on ultrasound velocity profiling and the distribution of sediment concentration." Geological Society, London, Special Publications 477 (2018a): SP477-19.

Nomura, S., Hitomi, J., Cesare, G. D., Murai, Y., Tasaka, Y., Takeda, Y., and Sakaguchi, H., "Instantaneous Flow Vector Measurement by a Pair of Ultrasound Doppler Instruments. ", Proc. of the 11th International Symposium on Ultrasonic Doppler Methods for Fluid Mechanics and Fluid Engineering 11th ISUD, September 5-7, (2018b), Berlin, Germany, pp. 119-122

Oehy, C. D., Cesare, G. D., Schleiss, A. J., "Effect of inclined jet screen on turbidity current." Journal of Hydraulic Research 48.1 (2010): 81-90.

Park, H. J., Tasaka, Y., Murai, Y., "Bubbly drag reduction accompanied by void wave generation inside turbulent boundary layers." Experiments in Fluids 59.11 (2018): 166.

Parker, G., Fukushima, Y., Pantin, H. M., "Self-accelerating turbidity currents." Journal of Fluid Mechanics 171 (1986): 145-181.

Pedocchi, F., García, M. H., "Acoustic measurement of suspended sediment concentration profiles in an oscillatory boundary layer." Continental Shelf Research 46 (2012): 87-95.

Piccirillo, P., van Atta, C. W., "The evolution of a uniformly sheared thermally stratified turbulent flow." Journal of Fluid Mechanics 334 (1997): 61-86.

Poelma, C., "Ultrasound imaging velocimetry: a review." Experiments in Fluids 58.1 (2017): 3. Povey, M. J., Ultrasonic techniques for fluids characterization. Elsevier, 1997.

Shi, X., et al. "Oil-gas-water three-phase flow characterization and velocity measurement based on time-frequency decomposition." International Journal of Multiphase Flow 111 (2019): 219-231.

Simpson, J. E., Britter, R. E. "The dynamics of the head of a gravity current advancing over a horizontal surface." Journal of Fluid Mechanics 94.3 (1979): 477-495.

Su, Q., Chao, T., Feng, D., "Mechanism modeling for phase fraction measurement with ultrasound attenuation in oil–water two-phase flow." Measurement Science and Technology28.3 (2017): 035304.

Symons, W. O., et al. "A new model for turbidity current behavior based on integration of flow monitoring and precision coring in a submarine canyon." Geology 45.4 (2017): 367-370.

Taishi, T., Kikura, H., Aritomi, M., "Effect of the measurement volume in turbulent pipe flow measurement by the ultrasonic velocity profile method (mean velocity profile and Reynolds stress measurement)." Experiments in Fluids 32.2 (2002): 188-196.

Takeda, Y., ed. Ultrasonic Doppler velocity profiler for fluid flow. Vol. 101. Springer Science & Business Media, 2012.

Theiler, Q., Franca, M. J. "Contained density currents with high volume of release." Sedimentology 63.6 (2016): 1820-1842.

Thorne, P. D., Hanes, D. M., "A review of acoustic measurement of small-scale sediment processes." Continental shelf research 22.4 (2002): 603-632.

Zheng, H., et al. "Real time multicomponent echo particle image velocimetry technique for opaque flow imaging." Applied physics letters 88.26 (2006): 261915.

Zordan, J., Schleiss, A. J., Franca, M. J. "Structure of a dense release produced by varying initial conditions." Environmental Fluid Mechanics (2018): 1-19.

Acknowledgements

この卒業論文を完成させるまでには，多くの皆様のご協力・ご支援をいただきました．ここで，感謝の意を示させていただきます．

村井祐一教授には，研究の奥深さや面白さ，難しさを徹底的に教えていただきました．研究に関する相談をさせていただく度に，当然のことながら「自分はまだまだ追いつけないな」と感じるばかりでした．それでも，何とかその差を縮めようともがいていた期間こそが，今振り返ってみると，私を大きく成長させてくれたと感じています．また，学会発表や海外インターンシップ，論文執筆等，まだまだ未熟な私に対して，数多くの貴重な経験をさせていただきありがとうございます．何事にも前向きに挑戦するという心構えを持てるようになりました．

田坂裕司准教授には，研究内容に関しては勿論のこと，英文執筆の際にも熱心な指導をしていただきました．研究発表や文章の添削をお願いした際には，私の稚拙な表現にもかかわらず，その意図をくみ取り，より良い表現方法を提案してくださいました．今でも私は，英語での表現には不安が残りますが，先生の熱心なご指導によって恐怖心を払拭することができました．また，研究面だけでなく，音楽のこと，スポーツのこと，自動車のこと，お子さんのこと等，多くのお話を聞かせていただき，楽しませていただきました．今後とも，先生の人間味溢れるご指導を続けていただきたいと思います．

朴炫珍助教授には，日常的に計測技術に関する相談に乗っていただきました．特に私の場合は，超音波計測がメインだったので，それに関して多くを教えていただきました．LFC のように実験系の研究室では，計測機器の取り扱いが，研究の善し悪しを決定づける大きな要素です．そのため，教えていただいた多くの超音波に関する知識は，実際に実験を行うときだけでなく，解析をする際にも大いに活用させていただきました．

山保敏幸技官には，実験装置の改良の際に大変お世話になりました．学部 4 年生の段階で既に，素晴らしい“シーソー装置”を作っていただきました．おかげさまで，その研究は論文にまとめることができ，修士課程のほとんどはスイスでの実験メインになりました．そのため，本格的な実験装置の作成をお願いする機会が無く，少し寂しい気持ちでした．それでも，装置に合わせた円筒形のアクリル部品が必要な旨をお話しした際には，翌日に完成させて持ってきてくださいました．やはり，山保さんの技術こそが LFC の研究実績を支える大きな核になっていることを改めて実感した瞬間でした．今後とも，お身体を大切になさってください．また，私が再び研究室を訪れた際には，自動車関係のお話をさせていただければ幸いです．

秘書の宮崎優衣さんには，学会参加や短期雇用関連，その他の事務的な手続き等で，大変お世話になりました．学部生の時には分からなかったのですが，修士課程になり，多くの出張等を経験させていただくうちに，宮崎さんの行う仕事の量とその早さに驚かずにはいられませんでした．また，そのようなお忙しい中でも，会議室の確保や書類関連の相談をさせ

ていただいた際には，快く対応していただきました．おかげさまで，私は研究活動に集中することができました．時には，野球✰話で楽しませていただきました．また改めて日ハム✰最新情報を聞かせていただく✰を楽しみにしております．

野村瞬博士には，修士課程✰研究で特にお世話になりました．「乱泥流」という研究テーマに対して，二人で試行錯誤しながら実験を行い，解析し，議論をして改善を繰り返した日々は忘れられません．ローザンヌ✰ EPFL ✰駅で初めて会った時から，気さくに話しかけてくださり，私も楽しい時間を多く共有させていただきました．研究だけでなく，旅行や食事にも幾度となく誘っていただき，大変感謝しております．二人で計測した乱泥流✰研究成果，並びに，JAMSTEC とLFC ✰共同研究✰より一層✰発展を願っております．また，ご家族✰皆さんにもご挨拶に伺わせていただく✰を楽しみにしております．

Dear Prof. De Cesare Giovanni: Thanks to your kind cooperation and many supports, I could have a good opportunity to study in LCH and to spend a precious time in Lausanne. You look so fun always and have a sense of humor, so I really enjoyed researching turbidity currents and discussing the analysis data with you. My dream is not only to bring the fruits of research on turbidity currents, but also to improve my English skill and to make a joke with you. I am looking forward to seeing you again.

武田靖教授には，乱泥流研究で大変お世話になりました．私が EPFL でインターンシップを行っている際には，何度もローザンヌまで足を運んでいただき，研究に関する議論をさせていただきました．「一見何も見えないと思い込んでいる実験動画でも，よく見ると流れ✰構造が表れている」というお言葉をいただいてから，私は計測手法だけでなく，流体力学に向き合いたいと思うようになりました．乱泥流プロジェクトは，工学部✰知見を活かして，新たな切り口から流れ✰内部構造を明らかにするという趣旨があります．そういった挑戦的なテーマに取り組むきっかけを与えてくださり，ありがとうございました．

LFC ✰先輩方，同期✰皆，後輩達にも感謝をしています．研究に行き詰った際には相談に乗ってくれたり，文章✰書き方や発表✰仕方に関して丁寧な指導をしてくれたりと，とても頼りがい✰ある LFC ✰先輩方を見てきました．いざ自分がそ✰立場になって，しっかりと先輩として✰役割を果たせたかは分かりません．それでも，一生懸命に研究に取り組む後輩達✰姿は，私✰研究に対するモチベーションに火をつけてくれましたし，研究以外✰場面でも，楽しい時間を共有してくれました．さらに，個性あふれる同期✰存在は，「ああ，自分は自分✰スタイルを貫けばいいんだな」と常に背中を押してくれるも✰でした．多様性に富んだ研究室✰皆さんと，流体力学✰テーマに対して真っ向から挑戦した期間は，私を大きく成長させてくれました．

こうして振り返ってみると，多く✰方々に支えられて，研究活動を行って来たことを実感しました．上に挙げた方々だけでなく，見守ってくれた家族であったり，意見交換をさせていただいた多く✰研究者✰方々であったり，関わってくださった皆さんに感謝✰気持ちを

示したいと思います．そして，LFC で培ったチャレンジ精神は，来年度からは社会人として発揮していけるよう頑張っていきたいと思います．皆さん✰益々✰ご活躍を祈っております．大変，お世話になりました．

人見純平

2019 年 2 月 21 日

List of publications

Journal
1. J. Hitomi, Y. Murai, H.J. Park, Y. Tasaka, “Ultrasound flow-monitoring and flow-metering of air–oil–water three-layer pipe flows”, IEEE Access, Vol.5, No.1 (2017), pp.15021-15029, DOI10.1109/ACCESS.2017.2724300 (Impact Factor =3.557)
International conference
1. Y. Murai, J. Hitomi, H.J. Park, Y. Tasaka, “Ultrasound Doppler velocimetry applied for air- oil-water three-phase pipe flows in slugging regime”, Proc. 11th Japan-U.S. Seminar on Two-Phase Flow Dynamics, (2017), pp.121-124.
2. S. Nomura, J. Hitomi, G. De Cesare, Y. Takeda, Y. Yamamoto, H. Sakaguchi, “Sediment mass movement of a particle-laden turbidity current based on ultrasound velocity profiling and the distribution of sediment concentration”, Geological Society London Special Publications, (2018), 477
3. S. Nomura, G. De Cesare, J. Hitomi, Y. Takeda, Y. Tasaka, Y. Murai, H. Sakaguchi, “Velocity and concentration distribution of saline density and turbidity current”, Proc. 5th IAHR Europe Congress - New Challenges in Hydraulic Research and Engineering (2018).
4. D. Noto, J. Hitomi, H.J. Park, Y. Tasaka, Y. Murai, “A proposal of a novel flow visualization method using rheoscopic fluid”, Proc. 8th ETH-HU Academic Exchange and Collaborative Research (AECoR8), (2018), pp.9-9.
5. J. Hitomi, S. Nomura, G. De Cesare, Y. Takeda, H. J., Y. Tasaka, Y. Murai, “Flow Monitoring of Particle-laden Flows Combining Ultrasonic Doppler and Echo Intensity Profiling Techniques”, Proc. of the 11th International Symposium on Ultrasonic Doppler Methods for Fluid Mechanics and Fluid Engineering 11th ISUD, (2018), pp. 75-78
6. S. Nomura, J. Hitomi, G. De Cesare, Y. Murai, Y. Tasaka, Y. Takeda, H. Sakaguchi, “Instantaneous Flow Vector Measurement by a Pair of Ultrasound Doppler Instruments”, Proc. of the 11th International Symposium on Ultrasonic Doppler Methods for Fluid Mechanics and Fluid Engineering 11th ISUD, (2018), pp. 119-122
7. J. Hitomi, S. Nomura, G. De Cesare, Y. Takeda, H.J. Park, Y. Tasaka, Y. Murai, “Comparison of flow behaviors between experimental turbidity currents and saline density currents combining ultrasonic and optical techniques”, Proc. 8th ETH-HU Academic Exchange and Collaborative Research (AECoR8), (2018), pp.10-10.
8. S. Nomura, J. Hitomi, G. De Cesare, Y. Takeda, “Steady state in experimentally produced turbidity current under a large amount of suspension supply”, EGU General Assembly 2019, (7–12 April, 2019).
Domestic conference (in Japan)
1. ○人見純平，朴炫珍，田坂裕司，村井祐一，“UVP を用いた管内水・油・空気層状流✰流量計測”，第 45 回可視化情報シンポジウム，（2017），Vol. 38, Suppl. No.1, A101, pp.1-5.
2. ○人見純平，野村瞬，De Cesare Giovanni，武田靖，阪口秀，朴炫珍，田坂裕司，村井祐一，“UVP を用いた乱泥流✰構造解析手法✰確立”，可視化情報全国講演会（室蘭 2017），（2017），No. OS4-1，pp.1-5.
3. ○澤田航希，人見純平，芳田泰基，朴炫珍，田坂裕司，村井祐一，“エコー強度情報を用いた超音波流速分布計✰多成分化”，日本機械学会北海道学生会第 47 回学生員卒業研究発表講演会，(2018), 317.
4. ○人見純平，野村瞬， De Cesare Giovanni，武田靖，朴炫珍，田坂裕司，村井祐一， “乱泥流内粒子群運動✰超音波による定量計測”，混相流シンポジウム 2018，（2018），No. 0059, pp.1-2.（ベストプレゼンテーション賞受賞）
5. ○李瑞雪，芳田泰基，人見純平，澤田航希，朴炫珍，田坂裕司，村井祐一, Stéphane Fischer，“ステンレス管内を流れる乳製品✰超音波診断”，日本機械学会第 96 期流体工学部門講演会，（2018），AM5-OS4-1, pp.1-3.
6. ○赤坂信太郎，人見純平，澤田航希，朴炫珍，田坂裕司，村井祐一，“超音波パルスエコグラフィーによる気液二相チャネル流✰高時間分解計測”，日本機械学会北海道学生会第 48 回学生員卒業研究発表講演会，(2019 年 3 月)

SPECIAL SECTION ON MULTIPHASE FLOW MEASUREMENT: TECHNIQUES AND APPLICATIONS

Received March 31, 2017, accepted June 6, 2017, date of publication July 11, 2017, date of current version August 22, 2017.

Digital Object Identifier 10.1109/ACCESS.2017.2724300

Ultrasound Flow-Monitoring and Flow-Metering of Air–Oil–Water Three-Layer Pipe Flows

JUNPEI HITOMI, YUICHI MURAI, HYUN JIN PARK, AND YUJI TASAKA

Laboratory for Flow Control, Division of Energy and Environmental Systems, Faculty of Engineering, Hokkaido University, Sapporo 060-8628, Japan

Corresponding author: Yuichi Murai (murai@eng.hokudai.ac.jp)

ABSTRACT The combined use of ultrasound pulse-echo intensity and Doppler shift frequency is examined as a means to measure strong unsteady three-phase pipe flows of a gas and two liquids. With air, oil, and water as components of the fluid media, particular attention is given to analyze ultrasound responses at the air–oil and oil–water interfaces. Reciprocating slugging is generated inside a 55-mm-diameter circular pipe, of which edges oscillate vertically at a controlled frequency. We use an ultrasound velocity profiler to obtain the 1-D cross-sectional distributions of the instantaneous flow velocity at the sampling rate of 60 Hz. All the measurements are realized by a single ultrasound transducer located outside the pipe. Measurement accuracy is validated using a high-speed camera coupled with particle image velocimetry that is synchronized with the profiler. The results demonstrate that the proposed technique works properly in sensing both interfaces as well as in-phase flow velocity distributions. In addition, multiphase volume flow rates for the constituents are obtained by velocity profile integration assuming vertical phase stratification in an approximation.

INDEX TERMS Doppler method, flow metering, multiphase flow, particle image velocimetry, pipe flow, pipe line, ultrasound

INTRODUCTIONUsing ultrasound Doppler information for flow metering is a non-invasive, first-principle-based measurement [1]. The volume flow rate is directly given by area integration of the spatial distribution of the velocity within a target plane. Distinct from ordinary indirect flow-metering devices, no cal- ibration is required preliminarily because of the capability of velocity distributions [2]. This feature enables the target of flow metering to be expanded widely such as for curved flows [3], non-Newtonian flows [4]–[6], and for multiphase flows [7]–[19]. In particular, the application to multiphase flow is challenging. Conventional approaches to flow meter- ing of multiphase flows are such that an individual phase needs to be sufficiently separated for the metering device to be effective. Alternatively, homogeneous mixing allows the devices to approximate the flow rate [20]. However, such phase-control operations require a change in pipeline sys- tems and invariably experiences a significant pressure loss. Therefore, for large systems such as in power and chemical plants, a non-invasive approach along with related technology to monitor directly internal multiphase flows is keenly waited to be developed.To handle the strong unsteady three dimensionality of multiphase pipe flows in the measurement section, sev- eral strategies are now being developed such as by electriccapacitance [21], [22], by optical rays [23], [24], and acoustic approaches. Among various non-invasive monitoring prin- ciples, ultrasound has multiple advantages in comparison with others. One is access to opaque fluids such as chem- ical solutions, fuel oils, and their suspensions. Another is spatial resolution, which suffices to monitor internal flow structures with a control of basic wavelength of ultrasound. As the sensors can be located outside the pipe, portability and easy-to-mount operation are further additional benefits to consider wide industrial applications. The known disad- vantage, conversely, is the damping of weak ultrasound pulses during propagation in fluid media. Users should also take care of acoustic noise from multiple reflections of pulses in the acoustic environment around the measurement section.In our previous series of developments, we found how to extract position of the fluid–fluid interface from the ultra- sound pulse waveform in echoes from multiphase flows. For gas–liquid combinations, three kinds of waveform analysis were proposed to obtain automated interfacial detection [15]. Based on the technique, we also developed an ultrasound void-fraction profiler [11] and ultrasound viscometry [17] for bubbly two-phase flow regimes. In the present work, we apply such ultrasonic sensing to three-phase pipe flows of gas–liquid–liquid stratifications. The demand for three-phase flow measurements have risen in the fuel pipe line industryVOLUME 5, 20172169-3536 © 2017 IEEE. Translations and content mining are permitted for academic research only.Personal use is also permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.15021
J. Hitomi et al.: Ultrasound Flow-Monitoring and Flow-Metering of Air–Oil–Water Three-Layer Pipe Flowswhere oil and water co-flow with a gas component. Such flows have a contrasting variety of flow patterns depending on the combination of three flow rates. The local flow structure must be monitored for stable continuous fuel transport as well as security from the risk of pulsatile behavior. Similar issues exist in chemical plants that handle complex multiphase flows consisting of phases of largely different properties.To meet these demands, we investigated specifically air– oil–water pipe flows to assess whether the ultrasound pulse technique is applicable. For the platform of experimentation, a seesaw type of oscillating three-phase pipe flow system was built (Fig. 1). Using this system, we obtained multiple interface echoes from a single sequence of ultrasound pulse emissions, and in-phase velocity profiling. From echo inten- sities and Doppler frequencies, we were able to examine flow metering to estimate the volume flow rates of each phase. The accuracy of the method is validated by particle image velocimetry enabling measurement error to be evaluated and the dynamics of three-phase pipe flows to be quantified.
FIGURE 1. Seesaw-driven multiphase pipe flow facility. Three phases are filled in the pipe with end plates on both edges, and oscillated by a single belt connected with a computer controlled linear slider. UVP andhigh-speed video imaging are conducted on the oscillation frame.
EXPERIMENTATION PLATFORM= =For the seesaw-driven facility (Fig. 1), the pipe diameter and length are D 55 mm, and L 1000 mm, respectively. To record the flow optically the pipe was made of transparent acrylic resin.Both ends of the pipe are connected to a single moving belt driven by a motor linear slider. The motion of this slider is controlled by a PC, which regulates the amplitude and frequency of the pipe ends. The maximum amplitude and frequency are 200 mm and 1.0 Hz. Air, oil, and water enter the pipe at the laboratory temperature of 25 ◦C. The test oil was Silicone oil of kinematic viscosity 10 cSt and density of 935 kg/m3 and is immiscible with water.
1. ULTRASOUND INSTRUMENTATIONThree kinds of measurement instrumentation were introduced to the platform providing ultrasound monitoring, backlight visualization, and laser sheet illumination.For ultrasound monitoring, an ultrasound transducer is set beneath the pipe at 15◦ to the pipe cross sectional plane.The transducer is submerged in a water jacket so that the propagation of the pulse into the pipe interior is easier as well as its return echo to the transducer. The water jacket is a cubic buffer tank with side length of 255 mm, sufficient to elim- inate ultrasound residuals from around as the measurement section is monitored at 2 kHz in pulse repetition frequency. The transducer emits and receives periodic ultrasound pulse of 4 MHz in basic frequency. The nominal intensity of the emitted pulse is 150 V whereas the echo intensity level is only of an order of a few millivolts. The effective diameter of the ultrasound pulse is 5 mm. The cycle number of the pulse is set to 4. Therefore, the spatial resolution in water is esti- mated to be a 5 mm-diameter disk of thickness 4[(1497 m/s)/ (4 106 Hz)]/2 0.75 mm, where 1497 m/s is the speed of sound in the experiment. The temporal resolution is given by the sampling rate of Doppler velocity profiling in the ultra- sound velocity profiling (UVP) operation, which is 60 Hz under present conditions.× =
2. OPTICAL FLOW VISUALIZATIONTo verify the ultrasound monitoring performance, two kinds of optical flow visualization were applied. One is backlight projection of the phase distribution [Fig. 2(a)], in which oil is dyed red, the dye being solvable in oil but insolvable in water. For water, aqueous green food dye was mixed in to identify the oil–water interface as contrast between red and green colors. Diffused white light of 650-W power illuminated the pipe from behind to enable a backlight color projection video image to be recorded. The camera was set up on the opposite side of the light source.
  FIGURE 2. Methods for visualizing the flow: (a) white backlighting of dyed liquids to see the projection phase distributions, and (b) laser sheet illumination of central plane where UVP measurement line is applied.The other is laser illumination of the oil and water phases for PIV measurements. For water, high-porous polymer par- ticles were used as the tracers. The particles have a density of 1010 kg/m3 and were 80 to 120 µm in diameter. For oil the tracers were water-repellent solid particles of den- sity 920 kg/m3 and peak diameter 110 µm. We observed that a small percentage of the particles adhered to the air– oil and the oil–water interfaces. Whereas the adhesion may change the interfacial property slightly, the Weber number (to be explained later) was kept high enough to have an
  J. Hitomi et al.: Ultrasound Flow-Monitoring and Flow-Metering of Air–Oil–Water Three-Layer Pipe Flowsinsignificant effect on the dynamic behavior of the three- phase flow. Moreover, the laser sheet scatters on the interfaces because of the adhesion, which assisted in viewing the two interfaces. The number density of tracer particles in each phase is guaranteed due to periodic wave-breaking gener- ated on both interfaces during the oscillatory motion of the pipe.= ===Fig. 3 shows how each phase behaves during a single 5-s-period cycle of the seesaw motion of the pipe. Left and right rows correspond with rightward and leftward flows driven by the pipe inclination. Note that the camera is mounted on the same oscillation frame so that consecutive images depict interfacial motions relative to the pipe wall. The pipe-bulk volume fractions of air, oil, and water are set to 1/3. The oil layer produces a wave-breaking front that rapidly propagates in the air–oil interface at t 1.0 s and t 3.5 s. Water accelerates slowly with a time lag of t 1.5 s and t 4.0 s after wave propagation in the upper oil phase. During this cycle, capillary waves move along the oil–water interface due to shear stress transfer through the interface. Slightly dark spots observed along both interfaces indicate wavy surfaces that accompany three-dimensional interfacial structures. Apart from these spots, the main flow is confirmed to be approximately two-dimensional.
  FIGURE 3. Cyclic behavior of the three-phase flow visualized using color projection in a single cycle of the pipe oscillation.Fig. 4 presents the time-line image of the phase distribu- tion, sampled from the video image. The sampling line is set vertically at the same position of the UVP measurement (to be described below). As the time-line image shows, the air–oil interface repeats a sharp rise and slow fall in every cycle whereas the oil–water interface fluctuates at a higher frequency but with small amplitude. With diverse combina- tions of volume ratios of air, oil, and water, many different types of interfacial behavior were observed. From these char- acteristics of the interfacial motion, ultrasound monitoring is examined.
  FIGURE 4. Time-line image of the three-phase flow at the central plane of the oscillating pipe, showing wavy passage of oil and water phases with different interfacial structures observed at the oscillation periodof T = 5 s.Fig. 5 presents the same time sequence of tracer particle images illuminated by a laser sheet. These images are pro- cessed by PIV software to verify the results from ultrasound measurements.
  FIGURE 5. Particle images for PIV measurements, in which fluid interfaces at the central plane are also visualized.
RESULTS AND DISCUSSIONTo analyze the data, for the oscillating three-phase pipe flow, dimensionless parameters are introduced:µjRej = ρjDU , j = {a, o, w}, (1),gD ∆ρ /ρ Frjk = U , jk = {a/o, o/w} (2)jk j=Wejk ρjUD, (3)σjk=where D and U denote the pipe diameter and bulk mean flow velocity, respectively. Subscripts a, o, and w denote the phase of air, oil, and water, respectively. The range of D treatable by the present ultrasound sensing is from 20 to 100 mm while we fix D 55 mm in the present demonstration. There are hence three Reynolds numbers—two Froude numbers, and two Weber numbers—giving a total of seven dimensionless parameters. In each definition, the characteristic speed of
J. Hitomi et al.: Ultrasound Flow-Monitoring and Flow-Metering of Air–Oil–Water Three-Layer Pipe Flows
the flow U is given by the pipe oscillation frequency F and amplitude H ,U = 2π FH (4)In the present range of test measurements, Re has order O(105) in water, O(104) in air, and O(103) in oil. This implies that the air–water interface is influenced by in-phase turbu- lence whereas the oil phase belongs to a transitional regime. Fr has order O(1) for air–water and O(10) for oil–water interfaces, evidencing the emergence of a wave-breaking front on both interfaces. The two Weber numbers are of order O(10) corresponding to a characteristic length scale of a few millimeters for capillary waves that may appear on the interfaces.Hence, the ultrasound sensing performance is assessed for all the above-mentioned flow elements stemming from within the three-phase flow, such as turbulence in each phase, density stratification, gravity-induced waves, shear-induced capillary waves, and periodic flow sandwiched by the two neighbor interfaces.
1. ECHO INTENSITY AND DOPPLER VELOCITY= (Prior to the experiment, we estimated the transmittance and reflectance of ultrasonic waves in the three-phase pipe flow including the pipe wall. Fig. 6 shows the result of the estimation with the three phases stratified vertically. With the ultrasound transducer set at the bottom of the pipe directed upward, echoes are received from six interfaces that exist between the transducer and the counter surface of the pipe. At each interface, formed between two media with acoustic impedances Z1, and Z2, the reflection ratio of ultrasound isFIGURE 6. Schematic of ultrasonic transmission via six interfaces along the straight path subject to different reflection ratio R and penetration of acoustic pressure AP.
  FIGURE 7. Position of ultrasound transducer for examination of orientation dependent signal characteristics. (a) Side view.(b) Cross-section view.With the estimation, we examined the signal quality of the echo by changing the angle of inclination of the trans-R Z2 − Z1 , Z1 = ρ1C1Z2 + Z1 Z1 = ρ1C1(5)ducer (Fig. 7). The results obtained (Fig. 8) were for equal volume ratios of three phases in the pipe, which oscillated at F = 0.2 Hz. The left and right columns present the echowhere ρ and C are density and speed of sound, respectively. The acoustic pressure A.P. of the ultrasound wave that prop- agates directly beyond multiple interfaces is.KA.P. = (1 − R1) (1 − R2) · · · = (1 − Rk ). (6)k=1=Here the acoustic pressure is normalized by the value of the initial pressure of the emitted wave. In the present configu- ration, the values of the refection ratios R and the acoustic pressure A.P. are obtained (Fig. 6). The acoustic pressure decreases to 0.30 in water, decreases further to 0.26 in oil, and almost disappears below 10−5 in air. The acoustic pressure of echoes that the transducer receives is further weakened by the square impact from A.P. because of shuttled propagation. Therefore we ignore the echo signal from the air phase as its acoustic pressure is below the sensing limit O(10−5). More- over, the echo from the air–oil and the oil–water interfaces are expected to be sufficiently strong as their pressure estimates are 0.07 and 0.01, respectively.intensity and Doppler velocity, respectively. Echo intensi- ties were originally recorded as electric voltages with both positive and negative signs of an order of millivolts, and subsequently normalized in a linear scale by the lowest limit of sensible voltage. Doppler velocities were computed from the Doppler frequency shift involved in echo wave- forms, the basic frequency of ultrasound, and the speed of sound. With the transducer inclined at ϕ < 60◦, the two interfaces are identifiable in the echo intensity plots. At ϕ 90, the interfaces no longer appear because from the for geometry the horizontal ultrasound pulses miss them. At ϕ >120, the echoes disappear completely because pulses encounter the first air phase blocking trans- mission to the liquid phases. From the Doppler velocity profiles, the fluid motion is properly captured with repeated cycles of oil and water phases. However, Doppler veloc- ities do not include any obvious information on interfa- cial position, but rather give a virtual velocity for the air phase. Next a method is described that eliminates this velocity.
  J. Hitomi et al.: Ultrasound Flow-Monitoring and Flow-Metering of Air–Oil–Water Three-Layer Pipe Flows
  FIGURE 9. Flow chart of UVP-data processing with interface detection to classify the domain into gas, oil, and water phases.FIGURE 8. Difference in echo intensity and Doppler velocity involved in the pulse waveforms taken in six different angles. (a) Echo intensity.(b) Doppler velocity.
2. METHOD OF INTERFACE DETECTIONWith a significant echo intensity confirmed from both inter- faces, we analyze the echo data to detect these interfaces automatically.The procedure (see outline in Fig. 9) involves first, apply- ing a median filter to reduce pixel noise associated with the time-space echo distribution. The size of a single pixel is defined by the temporal sampling frequency and the spatial sampling interval. Next, echo amplitudes are computed from the enveloping function, taking positive values over the entire domain. In this step, two sharp peaks appear in the spatial coordinates (middle graph of the figure) corresponding to the positions of the air–oil and oil–water interfaces. A threshold is given to detect the central coordinates of the intensity peaks. The threshold value can be determined reductively from the above-mentioned theoretical reflections. However, the reflections in actual situations are complicated through the influence of turbulence, beam divergence, and, capillary waves. Therefore, we determine the threshold from a recur- sive approach, selecting its value at which two positions of the strong echo are always detected. This threshold value is automatically computable from a statistical analysis of the echo data, and is relatively robust to the various flow behaviors taking place.The three-phase identification realized by the present inter- face detection is shown in Fig. 10. The figure involves
  FIGURE 10. Samples of interface detection: (a) optical visualization,(b) echo intensity distribution, (c) phase distribution, and (d) Doppler velocity distribution.individual time-line images of (a) the laser illumination of the two interfaces, (b) the echo intensity distribution, (c) the identification of the three-phase distribution, and (d) Doppler velocity profiles inside the oil and water phases. Regarding phase identification, some peaked errors remain at the air–oil interface in comparison with the color backlight visualiza- tion (see Fig. 4). These errors come from miss-detection of the air–oil interface due to sharp inclines in the local interface,
  J. Hitomi et al.: Ultrasound Flow-Monitoring and Flow-Metering of Air–Oil–Water Three-Layer Pipe Flowsfrom where no echoes are received. Nevertheless, the three- phase flow structure is captured with acceptable accuracy for general purposes such as for measurements of volume fractions, constituent volume flow rates, and spectral analysis of interface displacements.
3. FLOW VELOCITY PROFILING COMPARED WITH PIVThe flow velocity distribution in the oil and water are obtained from Doppler velocities using the ultrasound velocity pro- filing technique (UVP). There are two main concerns when UVP is applied to two kinds of co-flowing liquids in a single measurement section. One is the pulse refraction at the oil–water interface when inclined non-perpendicularly to the propagating direction of the pulse. The other is multi-dimensional flow events excited in the three-phase flow, which UVP cannot detect because velocity profiling is one dimensional. To quantify these features, particle image velocimetry (PIV) of the flow field is employed.= =From raw PIV data taken over a single cycle of the pipe oscillation, we extracted a time series of instantaneous flow velocity distributions. (Fig. 11); the magnitude of velocity vector is color coded. Here the pipe oscillation period is set at T 5.0 s (hence the frequency is F 0.2 Hz). The absence of local PIV data occurs in the result because the intensity of laser light saturates and incorrectly correlates the particle image in regions of rapid distortion in the fluid. After an adequate spatial interpolation is applied, a single vertical line along the same position of UVP is chosen to reproduce the time-line image of the velocity distribution (Fig. 12). Here the data for both UVP and PIV are the components of the flow velocity in the direction of ultrasound pulse propagation, i.e., 15◦ to the plane perpendicular to the axial direction of the pipe. From the figure, the UVP and PIV data exhibit strong similarities. Discrepancies in the data from the air– oil interface are attributable to absences of corresponding PIV data.A more quantitative comparison (Fig. 13) is obtained from===flow velocity fluctuations at mid-height in the oil phase (y 25 mm from the bottom) and in the water phase (y 15 mm). In the velocity computation from Doppler shift frequency, values for the speed of sound in oil and water are inputted, obtained already from information gleaned from the oil–water interface. In all tests under different flow condi- tions, the velocity magnitude of the oil phase from UVP is always larger than that from PIV. The ratio of the root-mean- square value of UVP to PIV ranges between 1.20 and 1.25. The high value is attributed to the mean dynamic inclination of the oil–water interface induced by slugging waves. In the water phase, such a bias is less than 10%. Note though that acoustic refraction on a flat oil–water interface, which has a refraction index of n 1.53, does not directly induce such bias error in the resultant velocity in the oil phase. The reason is simply that the pulse propagates both ways. In water-to-oil transmission, the pulse bends with refractive index n. From its own echo, it transmits a returning pulse propagating from oil to water that bends depending on the reciprocal index n−1 that
  FIGURE 11. Instantaneous PIV data obtained inside the two liquid phases at six different timing within a single oscillation.
  FIGURE 12. Comparison of fluid velocity data between UVP and PIV where PIV data are extracted on the same line of UVP.cancels out refraction in the oil phase. The present bias of the velocity magnitude is therefore attributed to curvature effects of the interface arising from the dynamics of multiphase flow.
4. CONSTITUENT FLOW RATESIn addressing our final objective, constituent volume flow rates of the three-phase flow were estimated from the measured information. The present technique provides a one- dimensional three-phase flow measurement that, in princi- ple, lacks information to correctly estimate the flow rates. An approximation model of the phase distribution is intro- duced for this estimation.In testing this approach, we modeled the local phase dis- tribution as a simple vertical stratification of the three lay- ers (Fig. 14). In this model, a single transducer is placed at the bottom of the pipe. The flow rate of each phase is computed using the area integration of the velocity distribution within
  J. Hitomi et al.: Ultrasound Flow-Monitoring and Flow-Metering of Air–Oil–Water Three-Layer Pipe Flows
  FIGURE 13. Flow velocity fluctuations in (a) the oil phase at y = 25 mm and (b) the water phase at y = 15 mm.∫the border of the target phase,FIGURE 14. Velocity profile integration: (a) side view, (b) cross-section view, and (c) definition of ux (y ).Q(t) =Aux(y, z)dydz, (7)= ===where ux is the flow velocity along the axial direction of the pipe, and (y,z) are the two-dimensional coordinates in its cross-sectional plane. A is the area over which the velocity distribution is integrated, and corresponds to the area of the target phase. The present ultrasound monitoring obtains the velocity distribution ux(y,z), the air–oil interfacial position yao(z), and the oil–water interfacial positionyow(z) on the central line at 0 < y < D, z 0. Conditions for perfect stratification of the three phases implies ux(y,z) ux(y,0), yao(z) yao(0), andyow(z) yow(0). This simplification leads to two formulae that give simultaneously the volume flow rates of both oil and water:( Qo(t) = . yao(0) ux(y, 0) · 2√y (D − y) · dy,
  FIGURE 15. Instantaneous volume flow rates measured bycross-sectional integration of axial velocity profile for each phase.monotonically, peaks, and gradually decreases, being simpler than the behavior of the constituent flow rates. This implies that the total flow rate behaves as a two-phase flow of gas and liquid [25]–[27]; its decomposition into three phases shows no functional similarity to the total liquid flow rate. A large peak in the total flow rate in the tail comes from a. √yow(0)0Qw(t) = yow(0) ux(y, 0) · 2 y (D − y) · dy.(8)three-phase flow instability, which takes place at the moment when the flow rate of water decelerates quickly and fallsThese quantities are computed using discrete numerical inte- gration based on the rectangular rule, and depend on the spa- tial sampling interval. Note that these two independent flow rates are obtained as a function of time at a given arbitrary sampling frequency of the ultrasound measurement.Fig. 15 shows the plots of the time-dependent flow rates for oil and water in the half-period of the pipe oscillation at sampling rate of 67 Hz. The oil phase moves earlier than the water phase and its flow rate peaks first. This is due to gravity waves in the oil phase propagating above the water phase. The water phase then flows and accelerates whereas the oilflow settles. When the water flow rate peaks, the oil flowbelow that of oil. Note that, in the present experimentation platform, the volume flow rate of air is given simply by Qa (Qo Qw) as the three phases are packed inside a single pipe with closed ends. In a long pipe without ends, the flow rate of the gas phase can be estimated using the equation of continuity for incompressible multiphase flow.= − +A quantitative assessment of the accuracy in flow rate measurements is possible as a cumulative flow volume is evaluated from the time-dependent constituent flow rates. The results for the cumulative flow volume for each phase (Fig. 16) were calculated usingmoves again which persists after the water flow stops. Large scattering of the data in the oil flow rate near the tail of the dis-V (t) =t∫t=0Q(t)dt.(9)tribution is attributed to a vertical fluctuation of the oil–water interface owing to shear-induced waves amplified. The total flow rate of oil and water phases (marked in black) increasesIn tests, we poured 400 mL of water and 300 mL of Silicone oil into the pipe. For the measurement system to demon- strate perfect accuracy, the cumulative flow volumes need
  J. Hitomi et al.: Ultrasound Flow-Monitoring and Flow-Metering of Air–Oil–Water Three-Layer Pipe Flows
  FIGURE 16. Measured cumulative flow volumes given by space-time integral of UVP-velocity data as 300 mL and 400 mL of oil and water were filled inside the pipe. Error bars indicate 10 % in relative error.to recover these same values. The measured results in the figure indicate that the error is 3% for water and 8% for oil. The error for the total liquid flow rate is 2%. By the same assessment for the other three-phase flow conditions, the maximum error has been found to be less than 10% for constituent flow rates and 5% for total flow rates. These evaluations are notable in the single slugging motion of the three-phase flow, and in principle, therefore, the flow rate over long-duration time averages, including unsteady flow, would yield better accuracy using the sampling law, i.e. the inverse square-root principle. For instance, 10% of the ran- dom error for a single event is comparable with 5% for four events, and 1% for 100 events. In applications, we assume users will choose time-average durations depending on their needs when applying the present measurement technique.
SUMMARY

An experimental technique for ultrasound monitoring of three-phase pipe flow was presented. We designed a see- saw type of pipe oscillation facility and an experimentation platform to control the internal three-phase behavior with arbitrary unsteadiness. For an in-depth analysis, multiple measurement instrumentations were mounted on the same frame. Our main focus on three-phase flow was periodic co-flowing stratified flow that accompanied multiple flow elements to evaluate the performance of ultrasound sensing. The flow features slugging wave fronts on the air–oil inter- face, shear-induced displacements on the oil–water interface, and turbulence in the tail of the stratified flow.

The following conclusions are drawn: (i) Ultrasound echoes are detectable from both the air–oil interface and the oil–water interface as long as the transmission occurs in the first of either liquid in the pipe; (ii) The flow velocity obtained from pulse Doppler signals is influenced by acous- tic refraction at the oil–water interface leaving a bias error dependence in the flow behavior; (iii) The identification of the fluid phase is realized by thresholding the echo intensity

profile. With the identification, the virtual Doppler veloc- ity distribution inherent in the air phase can be eliminated; and (iv) Constituent flow rates are obtained by spatial inte- gration of velocity distribution within each identified phase. The volume error is less than 10% for a single slugging event, which is equivalent to 1% in calculations of a 100-ensemble- averaged flow rate. The next step of development is to use multiple ultrasound measurement lines, which may improve the three-dimensional monitoring of three-phase flows.

Acknowledgments

The authors express their thanks to the technical sup- porting staff, Mr. Toshiyuki Sampo, Mr. Taiki Yoshida, Ms. Megumi Akashi (Hokkaido University) and collab- orative partner, Dr. Shinsuke Sato (Mitsubishi Heavy Industries, Ltd).

REFERENCES

Y. Inoue, H. Kikura, H. Murakawa, M. Aritomi, and M. Mori, ‘‘A study of ultrasonic propagation for ultrasonic flow rate measurement,’’ Flow Meas. Instrum., vol. 19, nos. 3–4, pp. 223–232, Jun./Aug. 2008.
Y. Takeda, ‘‘Velocity profile measurement by ultrasound Doppler shift method,’’ Int. J. Heat Fluid Flow, vol. 7, no. 4, pp. 313–318, 1986.
B. Beulen, A. C. Verkaik, N. Bijnens, M. Rutten, and F. van de Vosse, ‘‘Perpendicular ultrasound velocity measurement by 2D cross correlation of RF data—Part B: Volume flow estimation in curved vessels,’’ Experim. Fluids, vol. 49, no. 6, pp. 1219–1229, 2010.
M. J. W. Povey, Ultrasonic Techniques for Fluids Characterization. London, U.K.: Academic, 1997, pp. 11–44.
B. Ourlev and E. Windhab, ‘‘Novel ultrasound based time averaged flow mapping method for die entry visualization in flow of highly concentrated shear-thinning and shear-thickening suspensions,’’ Meas. Sci. Technol., vol. 14, no. 1, pp. 140–147, 2003.
T. Nakashima, T. Shiratori, Y. Murai, Y. Tasaka, Y. Takeda, andE. J. Windhab, ‘‘Viscoelastic responses of flow driven by a permeable disk investigated by ultrasound velocity profiling,’’ Flow Meas. Instrum., vol. 48, pp. 97–103, Apr. 2016.
T. Wang, J. Wang, F. Ren, and Y. Jin, ‘‘Application of Doppler ultra- sound velocimetry in multiphase flow,’’ Chem. Eng. J., vol. 92, nos. 1–3,pp. 111–122, 2003.
H. Murakawa, H. Kikura, and M. Aritomi, ‘‘Application of ultrasonic Doppler method for bubbly flow measurement using two ultrasonic fre- quencies,’’ Experim. Thermal Fluid Sci., vol. 29, no. 7, pp. 843–850, 2005.
Y. Murai, H. Fujii, Y. Tasaka, and Y. Takeda, ‘‘Turbulent bubbly chan-nel flow investigated by ultrasound velocity profiler,’’ J. Fluid Sci. Technol., vol. 1, no. 1, pp. 12–23, 2006.
M. R. Widyanto, M. B. Utomo, K. Kawamoto, B. Kusumoputro, and K. Hirota, ‘‘Local gas holdup measurement of a bubble column using SONIA-ultrasonic non-invasive method,’’ Sens. Actuators A, Phys., vol. 126, no. 2, pp. 447–454, 2006.
Y. Murai, S. Ohta, A. Shigetomi, Y. Tasaka, and Y. Takeda, ‘‘Development of an ultrasonic void fraction profiler,’’ Meas. Sci. Technol., vol. 20, no. 11,p. 114003, 2009.
S. R. G. A, Y. Murai, and Y. Takeda, ‘‘Chapter 1 ultrasound-based gas– liquid interface detection in gas–liquid two-phase flows,’’ Adv. Chem. Eng., vol. 37, pp. 1–27, Dec. 2009.
S. M. van der Meer et al., ‘‘Microbubble spectroscopy of ultrasound contrast agents,’’ J. Acoust. Soc. Amer., vol. 212, no. 1, pp. 648–656, 2007.
H. Murakawa, H. Kikura, and M. Aritomi, ‘‘Application of ultrasonic multi-wave method for two-phase bubbly and slug flows,’’ Flow Meas. Instrum., vol. 19, nos. 3–4, pp. 205–213, 2008.
Y. Murai, Y. Tasaka, Y. Nambu, and Y. Takeda, S. R. G. A, ‘‘Ultrasonic detection of moving interfaces in gas–liquid two-phase flow,’’ Flow Meas. Instrum., vol. 21, no. 3, pp. 356–366, 2010.
Y. Takeda, Y. Murai, and Y. Tasaka, ‘‘Ultrasonic multiphase flowme- ter, ultrasonic multiphase flow rate measurement program, and multi- phase flow rate measurement method using ultrasonic wave,’’ U.S. Patent 8 401 805 B2, Mar. 13, 2013.
J. Hitomi et al.: Ultrasound Flow-Monitoring and Flow-Metering of Air–Oil–Water Three-Layer Pipe Flows
Y. Tasaka, T. Kimura, and Y. Murai, ‘‘Estimating the effective viscosity of bubble suspensions in oscillatory shear flows by means of ultrasonic spinning rheometry,’’ Experim. Fluids, vol. 56, p. 1867, Jan. 2015.
H. J. Park, Y. Tasaka, and Y. Murai, ‘‘Ultrasonic pulse echography for bubbles traveling in the proximity of a wall,’’ Meas. Sci. Technol., vol. 26, no. 12, p. 125301, 2015.
M. M. F. Figueiredo, J. L. Goncalves, A. M. V. Nakashima, A. M. F. Fileti, and R. D. M. Carvalho, ‘‘The use of an ultrasonic technique and neural networks for identification of the flow pattern and measurement of the gas volume fraction in multiphase flows,’’ Experim. Thermal Fluid Sci., vol. 70, pp. 29–50, 2016.
G. Oddie and J. R. A. Pearson, ‘‘Flow-rate measurement in two-phase flow,’’ Annu. Rev. Fluid Mech., vol. 36, pp. 149–172, Jan. 2004.
A. J. Mwambela and G. A. Johansen, ‘‘Multiphase flow component volume fraction measurement: Experimental evaluation of entropic thresholding methods using an electrical capacitance tomography system,’’ Meas. Sci. Technol., vol. 12, no. 8, pp. 1092–1101, 2001.
W. Warsito and L.-S. Fan, ‘‘Measurement of real-time flow structures in gas–liquid and gas–liquid-solid flow systems using electrical capacitance tomography (ECT),’’ Chem. Eng. Sci., vol. 56, nos. 21–22, pp. 6455–6462,2001.
Y. Murai, Y. Matsumoto, and F. Yamamoto, ‘‘Three-dimensional measure- ment of void fraction in a bubble plume using statistic stereoscopic image processing,’’ Exp. Fluids, vol. 30, no. 1, pp. 11–21, 2001.
E. Hervieu, E. Jouet, and L. Desbat, ‘‘Development and validation of an X-ray tomography for two-phase flow,’’ Ann. New York Acad. Sci., vol. 972, no. 1, pp. 87–94, 2002.
Y. Murai, H. Oiwa, T. Sasaki, K. Kondou, S. Yoshikawa, andF. Yama-moto, ‘‘Backlight imaging tomography for gas–liquid two- phase flow in a helically coiled tube,’’ Meas. Sci. Tech., vol. 16, no. 7,pp. 1459–1468, 2005.
Y. Murai, S. Yoshikawa, S. Toda, M. Ishikawa, and F. Yamamoto, ‘‘Struc- ture of air–water two-phase flow in helically coiled tubes,’’ Nucl. Eng. Des., vol. 236, no. 1, pp. 94–106, 2006.
Y. Murai, K. Inaba, Y. Takeda, and F. Yamamoto, ‘‘Backlight imaging tomography for slug flows in straight and helical tubes,’’ Flow Meas. Instrum., vol. 18, nos. 5–6, pp. 223–229, 2007.

YUICHI MURAI received the Ph.D. degree in engineering from the University of Tokyo in 1996. He was involved in nuclear engineering with Fukui University and applied optics with the Impe- rial College London from 1995 to 2012. He has been a Full Professor with Hokkaido University since 2010. His research interests are dynamics of multiphase flows, measurement techniques of fluid flows, and renewable energy.

HYUN JIN PARK received the Ph.D. degree from Hokkaido University in 2016. He was with LG Electronics Inc. and the Institut de Mécanique des Fluides de Toulouse. He has been an Assistant Professor with Hokkaido University since 2016. His research interests are ship drag reduction, mul- tiphase flow, ultrasound technology, PIV, image processing, laser applications, and signal process- ing for various thermo-fluid physics and engineer- ing systems.

JUNPEI HITOMI received the degree from the Department of Mechanical Intelligence System Engineering, Hokkaido University, in 2017. He received the Hatakeyama Prize given for best stu- dents in the field of mechanical engineering and from the Japan Society of Mechanical Engineers in 2017. His main research is the development of ultrasound techniques for dynamic multiphase flow systems.

YUJI TASAKA received the Ph.D. degree from Hokkaido University in 2004. He started his pro- fessional research career with Hokkaido Uni- versity, as a Research Associate, and where he has been an Associate Professor since 2011. His research interests are flow instability and transition phenomena, as well as the development of mea- surement techniques.

11th International Symposium on Ultrasonic Doppler Methods for Fluid Mechanics and Fluid Engineering

Berlin, Germany (5.-7. September 2018)

Flow Monitoring of Particle-laden Flows Combining Ultrasonic Doppler and Echo Intensity Profiling Techniques

Jumpei Hitomi1,2, Shun Nomura2,3, Giovanni De Cesare2, Yasushi Takeda4, Hyun Jin Park1, Yuji Tasaka1, and Yuichi Murai1

1 Laboratory for Flow Control, Hokkaido University, N13-W8, Kita-ku, Sapporo, Hokkaido 060-8628, Japan

2 Ecole polytechnique fédérale de Lausanne EPFL, Plateforme de constructions hydrauliques PL-LCH, Station 18, CH-1015 Lausanne, Switzerland

3 Mathematical Science and Advanced Technology, JAMSTEC, Yokohama Institute of Earth Sciences, 3173-25 Showa-machi, Kanazawa-ku, Yokohama 236-001, Japan

4 Laboratory for Food Research Engineering, Swiss Federal Institute of Technology Zurich, Zurich 8092, Switzerland

The methodology to reveal inner structures of turbidity currents are required. Their behaviors have important roles to affect the sedimentation or transportation of fine particles. We have proposed a novel monitoring methodology for flows containing fine particles, combing Doppler velocity and echo information obtained from ultrasonic velocity profiler (UVP). In this study, we captured the relationship between the echo profiles and particle number distributions. To simplify turbidity currents, stirring flows in a cylindrical container were chosen. As suspended particles, quartz particles with 13.5 m in the central diameter was examined. In a certain range of quartz volume fraction , UVP can detect integral echo signals obeying Rayleigh scattering and Doppler-shift frequencies, although the quartz diameter is much smaller than ordinary tracer particles adopted for UVP. The echo distribution of   1% showed that the stirring flow causes the local particle number distributions. Moreover, the possibility to reconstruct particle number distributions from echo amplitude distributions was indicated.

Keywords: Echo intensity, Particle-laden fluid, Particle number density distribution, Solid-liquid two-phase flow, Rotational flow

IntroductionIn the hydraulic industry, liquid-solid two-phase flows which contain solid particles have important roles to utilize aquatic resources. For example, collecting minerals and fossil fuels under the sea and transportation or sedimentation mechanisms of solid particles in reservoirs have been paid attention to. An example of them and main topic of our research group is turbidity currents [1-3] which are gravity currents driven by the density difference between particle-laden fluid and ambient fluid. They have complicated flow structures due to turbulence and complex interaction caused between each particle. They show local mixture density fluctuation accompanying clusters and clouds of particles, which is hardly investigated by optical approaches and numerical simulations. Our research group, therefore, has proposed the direct monitoring methodology of such particle-laden flows using ultrasonic technique. Ultrasonic velocity profiler (UVP) [4] can be applied to opaque fluids and measure the spatio-temporal velocity distributions using frequency veering based on Doppler effects. In addition, echo signals which are reflected waves scattered by suspended particles give us beneficial information on particle number distributions. In past studies, scattering processes from clouds of fine particles are researched [5, 6]. Even echo signals obtained from UVP could be used to detect moving interfaces of multi- phase or multi-layer flows [7, 8]. In addition, they were applied to obtain the profiles of suspended sediment concentration [9] and to measure void fractions [10]. Combining two distributions of Doppler velocity andecho information, a novel monitoring methodology for turbidity currents expects to be established.The objective of this study is to reveal relationship between number density distribution of fine particles and echo information. Stirring flows in a cylindrical container were chosen to simplify particle-laden flows like turbidity currents. As suspension particles, quartz particles with order of 10 m in the central diameter are examined. This range of particle diameter is less than one-tenth the ultrasonic wavelength of central frequency4 MHz in water, so Rayleigh scattering, which is almost isotropic scattering, occurs. They are not used for UVP measurement basically, because the intensity of reflected waves from a quartz particle is too small to be detected by ultrasonic transducers. It was, therefore, another objective to elucidate the flow conditions and the volume fractions of the quartz particles that UVP can measure reasonable velocity distributions of particle-laden flows.
Experiments
1. Experimental setupThe experimental setup is shown in figure 1. The experiments were conducted in a cylindrical container with 100 mm outer diameter, and 3 mm thickness of lateral wall and 120 mm height. The cylinder is made of acrylic resin and filled with test fluids containing the fine particles. The cylinder does not have a lid and thus top of fluid layer is free surface. The cylinder was mounted inside the water jacket to keep uniform temperature and allow transmission of ultrasonic wave from the outside ofthe cylinder. The flow was driven by a stirrer with 60 rpms in the rotational speed, and a stirring bar dipped into the bottom of the cylinder.Velocity distribution and echo information were measured by UVP. The obtained ultrasonic signals reflected from clouds of particles were processed by UVP monitor model Duo (Met-Flow S.A.) into spatio-temporal distributions of Doppler velocity and echo information. An ultrasonic transducer with 4 MHz central frequency was fixed in the jacket with a horizontal displacement 10 mm from the center line. This off-axis measurement makes it possible to obtain the velocity component including not only radial but also azimuthal velocity component. In case of rigid body rotation, the obtained velocity components show uniform profile along the measurement line. The advantage of this method, therefore, is that flow structures can be imagined from obtained velocity distributions relatively easily and reconstructed assuming axisymmetric flow field. The transducer was set at 25 mm from the bottom of the cylinder to avoid the blind of ultrasonic propagation due to the free surface. Table 1 summarizes the setting parameters of UVP measurement. Especially, amplification is important parameter in this study. Both values of “gain start” and “gain end” were fixed at constant values not to amplify the echo value depending on the distance from the transducer.
  Figure 1: Schematic diagram of experimental setup and arrangement of the measurement line.
2. Test fluidIn this study, quartz (Quartz K13, Carlo Bernasconi AG) which is often used as laden particles of experimental turbidity currents was examined. Quartz has particle size distribution and the central diameter is 13.5 m, and density equals to 2650 kg/m3. It is predicted that Rayleigh scattering, which is almost isotropic scattering, occurs in the test fluids, because quartz diameter is less than about one-tenth the ultrasonic wavelength of central frequency4 MHz in water. The cylinder was filled with 500 mL of tap water and quartz was added based on each volume fraction . Total seven cases with different volume fractions  = 0, 0.01, 0.1, 1, 5, 10, and 15% were conducted.Table 1: Setting parameters of UVP.Central frequency 4 MHz
  Temporal resolution
  10
  ms
  Spatial resolution
  0.74
  mm
  Velocity resolution
  4.75
  mm/s
  Number of cycles
  4
  -
  Number of repetitions
  32
  -
  Amplification (gain)
  6-6
  -
Results and discussions
Figure 2: Distributions of time-averaged value of each volume fraction; (a) echo amplitude and (b) Doppler velocity distribution.Firstly, we tried to comprehend the trends of echo distribution measured in this setup. Figure 2 (a) shows the echo amplitude profiles obtained by UVP in each case. The measurements were conducted for 40 s after the flows reached adequate developed states, and the corresponding number of the velocity profiles for time- averaging is 4000. These echo amplitudes were calculated as the absolute values of difference frombackground ( = 0%) UVP echo values. In this figure, x axis indicates the distance from the center of the cylinder and the ultrasonic measurement direction is from left to right side. As the volume fraction becomes larger in the cases of  = 0.01, 0.1, and 1%, echo amplitude values are getting lager. Their echo amplitudes of these cases show large values on the opposite side from the ultrasonic transducer because of the reflection from the wall. In contrast, the reflected waves from the wall disappears under the conditions  ≥ 5% due to the attenuation of ultrasonic waves and the attenuation is getting larger depending on their volume fractions. Figure 2 (b) shows velocity distributions measured by UVP. These velocities show the components of the measurement direction that contain the azimuthal and radial velocity components. In the cases of 0.1 ≤  ≤ 5%, reasonable velocity distributions can be observed. In this study, the velocity profiles measured with usual tracer particles (DIAION HP20SS, Mitsubishi Chemical, diameter 60 – 150 m, density 1020 kg/m3) is defined as correct distribution, and “reasonable velocity distribution” means having good agreement with the correct distribution. The velocity profiles of 0.1 ≤  ≤ 5% have an accuracy that has a high cross-correlation value exceeding 0.95 for the correct distribution. The cases of  = 10 and 15%, however, do not show the reasonable velocity distributions. It seems that the attenuation of ultrasonic waves prevents the detection of echo signals by the ultrasonic transducer. In addition, the case of  = 0.01 shows a notable result. The velocity values of this case drop irregularly near the center of the cylinder. This phenomenon implies that the amount of quartz particles is much fewer than wall side due to the centrifugal force. As known from the velocity distribution, high velocity values can be observed near the center. As the distance from the center becomes longer, the velocity values are getting smaller. From these distributions, the existence of a free vortex is indicated. The density of quartz is about 2.65 times larger than that of water, so it is possible that particles are blown from the center to near the wall sides by the centrifugal force. These phenomena are indicated by echo distributions too, because their values also drop near the center under the cases of  = 0.1 and 1%.Basically, UVP measurements are applied to flows containing tracer particles with adequate size, density, and concentration to reflect ultrasonic waves. However, the concentrations of particles in these experiments are relatively larger in comparison with the flows of normal UVP measurements. The experimental results for0.1 ≤  ≤ 5% show reasonable velocity distributions. This range of particle concentration, therefore, might be useful to detect integral echo signals obeying Rayleigh scattering. That is why UVP can obtain velocity distribution, although quartz diameter is much smaller than tracer particles.To reveal the particle number distributions, a supplemental experiment was conducted. The mean volume fraction  was fixed at 1%, because the echo amplitude dramatically dropped near the centercomparing to other conditions and reasonable velocity distributions could be obtained. In this experiment, the stirrer worked in the same manner as previous experiments until 20 s. After 20 s later from start of the measurement, the stirrer was turned off. The results of echo amplitude and velocity distributions are shown in figure 3. The velocity distribution shows that the free vortex near the center disappeared immediately after turning the stirrer off. The echo amplitude distribution, moreover, dramatically changes. During the time from 20 to 40 s, homogeneous dispersion of the quartz particles can be assumed, because it had been completely mixed for the first 20 s. During the first 20 s, the low echo region exists near the center. In contrast, that band near the center disappears after 20 s later, and the echo values measured in the far region from the ultrasonic transducer becomes smaller than that of the first 20 s. As the result of this experiment, it is revealed that not only velocity but also echo distribution were changed dramatically depending on the motion of stirrer. It seems that the change of flow velocity induces the local particle number distribution.
Figure 3: Spatio-temporal distribution of (a) Echo amplitude and (b) Doppler velocity when stirrer was turned off at 20 s later.To evaluate the relationship between echo amplitude and particle number distribution, time-averaged values of echo amplitude of two cases were calculated. The first is averaged values of the first 15 s and the second is that of the last 15 s. These results are shown in figure 4. The figure 4 shows the characteristic distribution of echo values. The echo values measured in the first 20 s are larger than that of the last 20 s near the both walls, roughly |x| > 15. In contrast, the echo values measured in the first 20 s drop sharply near the center. This trend of the echo distribution seems to be related to the particle number distribution. That is to say, the stirring flow causes the free vortex and it blows particles from the center to near the wall sides due to the centrifugal force.The echo amplitude ratio at two conditions are shown in figure 5 (a). In this graph, the range of x axis shows from0 to 45 mm which is far side from the ultrasonic transducer. To capture the characteristics of the flow field, Shiratori et al. [11] had established the methodology to obtain two velocity components of azimuthal and radial direction assuming axisymmetric flow field. The two velocity components obtained by this methodology are shown in figure 5 (b). In these profiles, the azimuthal velocity is dominant in the cylinder. Range of the lower echo amplitude ratio shown in figure 5 (a), roughly 0 ≤ x ≤ 15 mm, corresponds to relatively higher azimuthal velocity range.
Figure 4: Time-averaged echo amplitude of each caseIt would be possible to reconstruct the particle number distributions from echo amplitude ratio. Echo intensity values are the resultant values after some interactions between ultrasonic waves and media including the particles and water, such as attenuation, scattering, and so on. Theoretical models and equations are required to establish the methodology and its applicable range, which can reconstruct particle number distributions.
Conclusions

This study revealed the relationship between echo information and number density distribution of fine particles which are much smaller than the wavelength of emitted ultrasonic wave. UVP was applied to particulate rotating flows driven by a stirrer in a cylindrical container. As suspension particles, quartz particles with order of 10 m in the diameter are examined. Although this range of particle diameter is smaller than ordinary tracer particles adopted for UVP, reasonable velocity distributions could be obtained under the condition where the volume faction of particles  is from 0.1 to 5% due to the detection of the integral echo signals which obeys Rayleigh scattering. Paying attention to the echo distribution observed at  = 1%, the trend of particle number distribution in the cylinder was newly explored. As the result of the experiment, it was confirmed that the range of lower echo amplitude ratio corresponded to the range such that azimuthal velocity is relatively large. It is therefore inferred that the free vortex containing the dominant azimuthal velocity components relates to the local particle number density, which can be predicted from echo amplitude distribution. As the future work, the

theoretical models and equations describing the model are required to establish comprehensive methodology to reconstruct particle number distributions from echo values.

Figure 5: (a) Profiles of echo amplitude ratio at two conditions and (b) profiles of the azimuthal and radial velocity component.

References

Altinaker MS et al.: Flow structure in turbidity currents, J. Hydraul. Res., 34 (1996), 713-718.
Simpson, JE: Gravity currents in the environment and the laboraotory, Cambridge University Press, London, (1997).
De Cesare G et al.: Experiments on turbidity currents influenced by solid and permeable obstacles and water jet screens, ISUD6, (2008).
Takeda Y: Ultrasonic Doppler velocity profiler for fluid flow, Springer, (2012).
Thorne PD & Hanes DM: A review of acoustic measurement of small-scale sediment processes, Cont. Shelf. Res., 22 (2002), 603−632.
Lee TH & Hanes DM: Direct inversion method to measure the concentration profile of suspended particles using backscattered sound, J. Geophys. Res., 100 (C2) (1995), 2649−2657.
Murai Y et al.: Ultrasonic detection of moving interfaces in gas–liquid two-phase flow, Flow Meas. Instrum., 21 (2010), 356-366
Hitomi J et al.: Ultrasound flow-monitoring and flow- metering of air–oil–water three-layer pipe flows, IEEE access, 5 (2017), 15021-15029.
Pedocchi F & García MH: Acoustic measurement of suspended sediment concentration profiles in an oscillatory boundary layer, Cont. Shelf. Res., 46 (2012), 87−95.
Murai Y et al.: Development of an ultrasonic void fraction profiler, Meas. Sci. Technol., 20(11), (2009), 114003.
Shiratori et al.: Development of ultrasonic visualizer for capturing the characteristics of viscoelastic fluids, J. Vis., 16 (2013), 275–286.

8th ETH-HU Academic Exchange and Collaborative Research (AECoR8)

Comparison of flow behaviors between experimental turbidity currents and saline density currents combining ultrasonic and optical techniques

Jumpei Hitomi*, Shun Nomura **, Giovanni De Cesare***, Yasushi Takeda****, Hyun Jin Park*, Yuji Tasaka*, Yuichi Murai*

*Laboratory for Flow Control, Faculty of Engineering, Hokkaido University

** Mathematical Science and Advanced Technology, JAMSTEC

*** Laboratoire de Constructions Hydrauliques, École Polytechnique Fédérale de Lausanne (EPFL)

**** Food Process Engineering Laboratory, Swiss Federal Institute of Technology Zürich (ETH)

hitomi@ring-me.eng.hokudai.ac.jp

Turbidity currents [1], which are driven by density difference between particle-suspended fluid and ambient fluid, have important roles relating to the transportation and sedimentation of fine particles under seas and rivers. They are reported to travel much longer than distance expected from theories and numerical simulations despite containing numerous particles with heavier density than water, the distance is likely to reach a few thousand kilometers. Such a long-range propagation mechanism is not revealed completely yet due to the complicated flow structures and some interactions of particle- particle or particle-fluid. Our research group focuses on the experimental study of the turbidity currents. In this report, we highlight characteristics of turbidity currents by comparing with saline density currents. Those currents were generated by the lock-exchange method using a rectangular container with 4548 mm in overall length, 210 mm in height, and 143 mm in width. The initial densities of the test fluids in all cases were fixed at 1008 kg/m3. Total 11 currents were examined, and 8 of them were turbidity currents and the others were density currents. Their macroscopic motions were evaluated from instantaneous images taken by a video camera, and inner velocity distributions were measured by ultrasonic velocity profiler (UVP) [2] that can be applied to opaque fluids such a particle-suspended fluid. As shown in Fig.2, a significant

difference relating to the front velocity could be found depending on the presence or absence of fine particles. In the presentation, quantitative elucidation of long-range propagation mechanism of turbidity current will be explained from spatiotemporal velocity distributions obtained by UVP.

Fig.1 Experimental setup

Fig.2 Displacement of head position measured from each experimental image

References

B. Kneller, C. Buckee, The structure and fluid mechanics of turbidity currents: a review of some recent studies and their geological implications. Sedimentology, Vol.47, pp. 62-94 (2000)
Y. Takeda, Ultrasonic Doppler velocity profiler for fluid flow, Springer, (2012).

UVP を用いた管内水・油・空気層状流の流量計測

人見純平○ (北大），朴炫珍(北大），田坂裕司(北大），村井祐一(北大）

UVP flow-metering of air–oil–water three-layer pipe flows

Jumpei HITOMI, HyunJin PARK, Yuji TASAKA and Yuichi MURAI

ABSTRACT

We have proposed a novel multi-phase flowmeter using ultrasonic velocity profiler (UVP) for quantitative measurement of opaque complex flows in fuel pipeline systems arranged at bottom of the sea. Echo intensity obtained from UVP makes it possible to detect not only gas–liquid but also liquid–liquid interfaces at the same time. UVP flowmeter, therefore, can be extended from single-phase flows to multi-phase flows. Using both of echo intensity and Doppler shift frequency, instantaneous interface positions and velocity distribution are measured, so volume flow rates of each sampling time can be obtained. As a result of comparison with the results of optical visualization, it was revealed that the difference of speed of sound in each phase and refraction of ultrasound on interfaces contribute largely to the calculation of measurement positions and velocities. This paper also proposed the correction method for these and validity of UVP multi-phase flowmeter were mentioned.

Keywords: Visualization, Ultrasound, Flow rate, UVP, Echo intensity

Temporal resolution	10	ms
Spatial resolution	0.74	mm
Velocity resolution	4.75	mm/s
Number of cycles	4	-
Number of repetitions	32	-
Amplification (gain)	6-6	-

Master thesisInner flow structures of turbidity currents based on applied ultrasonic techniques​

Download the full PDF ➡

Read the full paper ⬇

Inner flow structures of turbidity currents based on applied ultrasonic techniques

Jumpei HITOMI

Laboratory for Flow Control,

Division of Energy and Environmental Systems, Graduate School of Engineering,

Hokkaido University.

1st February 2019

‌Experimental method

‌Results and discussions

‌Conclusion

‌Bibliography

Master thesisInner flow structures of turbidity currents based on applied ultrasonic techniques