flowregimeidentiﬁcationoftwo-phaseliquid–liquidupﬂow ......flow of a mixture of two immiscible...

Chemical Engineering Science 61 (2006) 1500–1515www.elsevier.com/locate/ces

Flow regime identification of two-phase liquid–liquid upflowthrough vertical pipe

A.K. Janaa, G. Dasa, P.K. Dasb,∗aDepartment of Chemical Engineering, Indian Institute of Technology, Kharagpur-721302, India

bDepartment of Mechanical Engineering, Indian Institute of Technology, Kharagpur-721302, India

Received 7 July 2005; received in revised form 24 August 2005; accepted 2 September 2005Available online 2 November 2005

Abstract

The present work has attempted to identify the flow patterns during liquid–liquid two phase flow through a vertical pipe. Dyed keroseneand water have been selected as the test fluids. The measurements have been made for phase velocities varying from 0.05 to 1.5 m/s for boththe liquids. The conductivity probe technique has been adopted and three different probe designs have been used to identify the patterns underdifferent flow conditions. A parallel wire type probe traversing the entire cross-section along a diametral plane has indicated the existence ofbubbly flow at low phase flow rates and dispersed bubbly flow at high velocities of water. Apart from the visual appearance of the signals,different statistical analysis namely the probability density function and wavelet analysis have been performed for a better appraisal of the flowsituation. The information in the PDFs have been quantified by means of the statistical moments. The existence of the core-annular flow at highkerosene and low water velocities has been confirmed from measurements using a different probe design. The intermediate region between thebubbly and annular flow patterns is characterized by a random distribution of the two liquids with continually changing interface between them.This has been named as the churn turbulent flow pattern. The information thus obtained has been represented in the form of a flow pattern map.� 2005 Elsevier Ltd. All rights reserved.

Keywords: Multiphase flow; Hydrodynamics; Vertical pipe; Dispersion; Annular flow; Instrumentation

1. Introduction

Flow of a mixture of two immiscible liquids occurs in manyindustrial processes and in the petroleum industry in particu-lar, where oil and water are often produced and transported to-gether. During their cocurrent flow in a pipe, the deformableinterfaces of the two fluids can assume a variety of charac-teristic configurations, which can be classified into differentflow patterns or flow regimes. The flow patterns cannot be pre-dicted from the independent variables of the system such as thephase flow rates and their physical properties in a straightfor-ward manner. An identical observation has also been reportedfor gas–liquid flows by the past researchers. They have mostlypresented observations of flow patterns in the form of a plottermed as the flow pattern map where the most commonly usedaxes represent the superficial velocities of the two phases.

∗ Corresponding author. Tel.: +91 3222 282916; fax: +91 3222 255303.E-mail address: [email protected] (P.K. Das).

0009-2509/$ - see front matter � 2005 Elsevier Ltd. All rights reserved.doi:10.1016/j.ces.2005.09.001

The interest on flow patterns arises due to the fact that the hy-drodynamics of flow depends on the interfacial configurations.When these distributions are taken into account, more accu-rate models can be developed for two-phase flows. Clearly theflow patterns would be expected to vary with (for a given pipediameter and orientation) the velocities, the volume fractionsand physical properties (density and viscosity) of the respec-tive phases. A further parameter which is likely to be importantfor liquid–liquid cases is the wetting characteristics of the tubewall. Wetting effects can be important in gas–liquid flows forhydrophobic channel walls but are not usually taken into ac-count. It has been reported that the manner in which the phasesare introduced into the conduit also influences the prevailingpattern.

Experimental studies on flow pattern maps during horizon-tal oil–water flows were successfully done by Russell et al.(1959), Charles et al. (1961), Hasson et al. (1970), Guzhovet al. (1973), Arirachakaran et al. (1989), Trallero (1995), Valleand Kvandal (1995) and Nadler and Mewes (1995). In addition

http://www.elsevier.com/locate/cesmailto:[email protected]

A.K. Jana et al. / Chemical Engineering Science 61 (2006) 1500–1515 1501

to the experimental studies on flow regimes, criteria of flowpattern transition have been given by Brauner and Maron (1992)(for stratified, stratified-dispersed, annular, slug and dispersedflow regimes), Brauner (2001) (for dispersed flow boundary)and Brauner and Ullmann (2002) (for oil-in-water dispersionand water-in-oil dispersion).

Experimental studies of oil–water flow in inclined pipes werereported by Mukherjee et al. (1981), Vigneaux et al. (1988),Flores et al. (1998). Mukherjee et al. (1981) measured pressureloss and water holdup for oil–water flow in 1.5 in diameter pipewith inclination angle varying from ±30◦ to ±90◦ from thehorizontal. Vigneaux et al. (1988) measured the distribution ofthe water volume fraction across a pipe section during oil–waterflow. They used local high frequency impedance probes in a20 cm ID pipe at mean velocities between 2.7 and 35 cm/s, atdeviation angle between 0◦ and 65◦ from vertical, and at watervolume fraction between 30% and 100%. Flores et al. (1998)carried out theoretical as well as experimental investigations ofoil water flow in vertical and inclined pipes. The tests coveredinclination angles of 90◦, 75◦, 60◦ and 45◦ from the horizon-tal. They reported the holdup and pressure drop behaviors tobe strongly affected by oil–water flow patterns and inclinationangle.

One of the earliest experimental studies on liquid–liquidtwo-phase flow through vertical pipes dates back to Govieret al. (1961). The authors studied pressure drop and holdupusing three different oils. Brown and Govier (1961) studiedpressure drop, bubble velocity and bubble size distribution inoil–water vertical flow using high-speed photography. Hassonet al. (1974) have presented analytical as well as experimentalstudies on liquid–liquid annular flow through vertical conduits.The analysis has considered laminar flow in both the liquidsand described the flow field as a superposition of an undis-turbed field and a disturbance field. Their experimental resultshave validated the analysis and defined its range of applica-bility. The authors have further shown the Lockhart–Martinellimodel to be inadequate for liquid–liquid annular flows. Farrarand Bruun (1996) applied a hot film anemometer based tech-nique in the study of kerosene–water two-phase flow in thebubbly, spherical cap bubble and churn flow regimes. The au-thors presented radial bubble volume fraction profile, bubblecut chord length profile, bubble mean velocity profile and tur-bulent intensity profile. Hamad et al. (1997, 2000) developedoptical probe systems and studied kerosene–water two-phaseflow through vertical pipe.

Some studies have also been performed to study the effectof pipe material on the hydrodynamics of liquid–liquid flow.Angeli and Hewitt (1998) performed experiments in stainlesssteel and acrylic tube and proposed that the material of the tubewall can strongly influence the pressure gradient during two-phase liquid–liquid horizontal flow. Pressure gradients underall conditions were higher in the steel than in the acrylic tubefor the same mixture velocities and flow volume fractions, thedifference being greater than what would be expected fromthe difference in the wall roughness. Angeli and Hewitt (2000)carried out experiments in horizontal stainless steel and acrylictube and concluded that in the stainless steel tube the propensity

for dispersion was greatly increased; in the acrylic tube oiltended to be the continuous phase for a wider range of flowconditions than in the steel tube. Ioannou et al. (2005) studiedphase inversion in steel and acrylic pipes and concluded thatthe pressure gradient peak around phase inversion is sharperand larger in the acrylic pipe as compared to the steel one withthe same diameter.

The above survey shows that unlike gas–liquid flows, the flowpatterns in liquid–liquid systems and consequently the flow pat-tern map has not yet been standardized. Moreover, the majorityof the studies in liquid–liquid flows are confined to horizon-tal pipes. This has motivated the present study to perform anindepth investigation of the flow patterns during liquid–liquidup flow through vertical conduits and to develop an objec-tive method for identifying the transition between subsequentpatterns.

2. Experimental setup and procedure

The schematic diagram of the experimental set up designedand fabricated to investigate vertical up flow of kerosene–watermixture is shown in Fig. 1. It consists of a test rig and acces-sories namely water tank, kerosene tank, kerosene–water sep-arator, two centrifugal pumps and measuring equipment. Thetest section comprises of a vertical transparent acrylic resin(perspex) tube of 0.0254 m diameter and 1.4 m length. Acrylicresin was selected as the material of construction to facilitatevisual observation of the flow phenomena. An entry length of2.0 m is provided to ensure fully developed flow. After the testsection, there is an exit length of 0.60 m to avoid any flow dis-turbance in the test rig. In the test section, a glass view box(VB) of 0.30 m length is attached for photography. The test flu-ids are water and dyed kerosene. Blue dyed kerosene has beenused in the experiments for better visualization of the flow phe-nomena. They are pumped through pumps (P1 and P2) fromtheir respective storage tanks. The flow rates are metered bypreviously calibrated rotameters. The two liquids are then in-troduced by a T arrangement at the entry where water and oilenter from the vertical and horizontal directions, respectively.After the test section, the kerosene–water mixture enters a sep-arator. Both liquids are then directed into their respective stor-age tanks after getting separated by gravity.

The superficial velocities of both water and kerosene havebeen varied from 0.05 to 1.5 m/s. The experiments are carriedout by increasing kerosene velocity at a constant water veloc-ity. The water velocity is then changed and the readings arerepeated. Next, all the measurements are carried out in the re-versed way i.e., keeping kerosene superficial velocity constant,the water superficial velocity has been increased continuouslywithin the range to study the existence of hysteresis. The phys-ical properties of water and kerosene are given in Table 1.

2.1. Measurement technique

Numerous techniques exist for estimation of flow patternsin gas–liquid flow. The most common way to identify the dif-ferent flow patterns is to observe the flow in a transparent

1502 A.K. Jana et al. / Chemical Engineering Science 61 (2006) 1500–1515

Fig. 1. Schematic diagram of experimental setup.

Table 1Physical properties of water and kerosene (at 30 ◦C and atmospheric pressure)

Fluid Density Viscosity Surface tension(kg/m3) (N s/m2) (N/m)

Kerosene 792 0.00137 0.027Water 1000 0.001 0.072

channel or through a transparent window on the pipe wall.As an extension to visual observation, photographic or video-graphic recording have also been widely used. For two-phasemixtures moving very rapidly, high speed photography orvideography is necessary. However, even high speed pho-tography/videography is often not sufficient to give a cleardelineation of the flow pattern, since complex interfacialstructures result in multiple reflection and refraction that ob-scure the view, especially in the core of the pipe. Besides,photographic/videographic techniques can be misleading es-pecially close to the flow pattern boundaries, where the visual

differences between two patterns can be very small. As a re-sult, in gas–liquid flows, a variety of other objective techniquesnamely the photon attenuation technique (Jones and Zuber,1975), the impedance probe method (Sun et al., 2004), the hotfilm anemometry (Serizawa et al., 1975; Wang et al., 1987;Farrar and Bruun, 1989; Liu and Bankoff, 1993; Al-Deenet al., 1998; Hamad and Brunn, 2000; Rensen et al., 2005,etc.) have been used to supplement the visual observations.Among them, the conductivity/impedance probe method hasbeen reported to be very effective and useful. It works on theprinciple that the electrical impedance of a two-phase mixtureis a function of concentration. Its popularity arises due to itslow cost and almost instantaneous response. Probes of differ-ent geometries such as point electrode (Serizawa et al., 1975),arc electrode (Cheng et al., 2002), ring electrode (Andreussiand Bendiksen, 1989), wire electrode (Miya et al., 1971), stripelectrode (Das et al., 2000) have been used widely for intru-sive and nonintrusive applications as well as point and globalmeasurements. Apart from flow pattern identification, theyhave been adopted for measurement of different hydrodynamic


parameters namely bubble size, frequency and velocity in bub-bly flow, liquid film thickness in annular flow etc.

In liquid–liquid flow on the contrary, most of the investi-gators have used visual observation and photography relatedtechniques (Russell et al., 1959; Charles et al., 1961; Hassonet al., 1970; Pacek et al., 1994). Hamad et al. (1997, 2000)have used intrusive optical probes for measurement of localvolume fraction profile, drop velocity and drop size measure-ments in kerosene–water bubbly flow. Simmons and Azzopardi(2001) have utilized a laser diffraction technique for drop sizestudy in kerosene and aqueous potassium carbonate solutionfor both vertical upflow and horizontal pipes. A few studieshave adopted the conductivity/impedance probe. Vigneauxet al. (1988) have used the high frequency (1 GHz) impedanceprobe consisting of 0.5 mm diameter tip. The measurementsare based on the difference between the impedance valuewhen the probe tip is immersed in oil or water. Angeli andHewitt (2000) have determined local phase fraction in oil–waterflow through horizontal pipe with a high frequency impedanceprobe and detected the continuous phase in dispersed flowswith a needle type conductivity probe. Jin et al. (2003) havecharacterized the flow patterns in oil/water two-phase flow invertical pipe by the analysis of fractal, chaos and Kolmogoroventropy of the conductance time series fluctuating signals. Theprobe consisted of four ring type stainless steel measuringelectrodes flush mounted on the inside wall of the insulatingpipe. Ioannou et al. (2004) applied a conductivity probe tomonitor the change of the continuous phase and an impedanceprobe to measure the volume fraction distribution near thephase inversion point in a horizontal pipe. The conductivityprobe consisted of two 0.5 mm diameter copper wires, placedvertically in a pipe cross-section 10 mm apart and indicated thephase continuity at a particular cross-section. The impedanceprobe consisted of a coaxial two-electrode tip which allowedmeasurements of local volume fraction at a point in the pipecross-section. Lovick and Angeli (2004) identified the dual con-tinuous flow pattern boundaries with the use of impedance andconductivity probes described by Ioannou et al. (2004). Ioannouet al. (2005) applied impedance ring probe to detect phase in-version in liquid–liquid dispersed flow. The probe consisted ofthree impedance ring pairs placed along the pipe to investigatewhether phase inversion happens simultaneously in the wholetest section.

From the above survey, it is evident that the conductivityprobe method is well established in gas–liquid flows and hasbeen applied with limited success for liquid–liquid cases. Mostof the researchers have used an impedance probe along withthe conductivity technique and they have reported that the con-ductivity probe technique is effective primarily when water isthe continuous phase. The sensitivity of the probe is also lessin liquid–liquid systems due to the lower difference in con-ductivity between the two liquids as compared to that betweenair and water. An additional difficulty also arises when the oilsticks to the probe surface and alters its response.

In the present study, the flow patterns are initially noted byvisual observations and photographic techniques. It was notedto be effective at low phase velocities but failed to identify the

Tube wall

Thin wires

Fig. 2. Parallel wire conductivity probe.

distribution at high flow rates. The conductivity probe tech-nique has, therefore, been adopted to identify the flow patternsduring the simultaneous flow of water and kerosene through avertical pipe. The design of the probe should be such that itwould be in close proximity with the two-phase mixture overthe entire cross-section while ensuring minimum obstruction ofthe flow passage. With this consideration, a parallel wire typeconductivity probe, which traverses the cross-section along adiametral plane, has been designed in the present study. It con-sists of two 0.10 mm diameter stainless steel wire stretchedhorizontally in the pipe cross-section parallel to each other andlocated 10 mm apart (Fig. 2). It has been installed at a dis-tance of 2.5 m from the entry section. The Wheastone bridgeprinciple has been employed to enhance the sensitivity of theprobe. The probe is connected in parallel to one of the resis-tance arms of the bridge and the other arms are so adjustedthat the bridge is balanced when the probe is totally dipped inwater. The unbalanced voltage of the bridge when the probe ispartially submerged in water is amplified through a differentialamplifier. The amplifier output is connected to a precision rec-tifier, which is capable of rectifying a signal as low as 10 �V.The DC output of the rectifier is recorded continuously in acomputer via Data acquisition system. An Agilent 34970A Dataacquisition/Switch unit with 16 channels has been adopted forthis purpose. It records data with a frequency of 25 Hz over aperiod of 100 s. The probe output provides an insight into thechordal average phase distribution across the pipe diameter.

Initially the voltage signal is recorded for different veloci-ties of single phase water and kerosene, respectively. This hasyielded a steady output at a single voltage for each liquid. Thevalue is higher for water as compared to kerosene and it is in-dependent of the liquid velocity. During the simultaneous flowof the two fluids, the voltage signals are normalized by Vmax,the voltage value obtained when only water passes through thepipe in order to facilitate a comparative study of the randomsignals under different flow conditions.

2.2. Method of analysis

Apart from visual observations of the random signals, differ-ent statistical analysis of the normalized time series data have


been carried out to get better information about the phase dis-tribution in the pipe. The probability density function (PDF)and the wavelet analysis have been adopted for this purpose.

2.3. PDF analysis

The PDF is a well established technique for time series anal-ysis of random signals. It gives the time-averaged histogramdepicting the distribution of amplitudes of the signal. Severalresearchers (Jones and Delhaye, 1976; Jones and Zuber, 1975;Vince and Lahey, 1982; Li, 2002) have adopted the tech-nique for identifying flow patterns during gas–liquid flows.The details of the analysis have been provided in Jones andDelhaye (1976). If a random variable X has a cumulativedistribution function F(x) which is differentiable, the PDFis defined as f (x) = dF/dx. The probability of observing Xin the interval x�X < x + dx is then f (x) dx. For severalvariables X1, X2, . . . , Xn the PDF is

f (x1, x2, . . . , xn) = �nF (x1, x2, . . . , xn)

(�x1�x2 . . . �xn).

A better method of quantification of the PDF curves can beaccomplished by means of the statistical moments namelymedian, variance, skewness and kurtosis. The skewness (thirdmoment) and kurtosis (fourth moment) are nondimensionalmoments, in contrast to the median and standard deviation(�) which has the same dimension as the measured quantities.The skewness characterizes the degree of asymmetry of thedistribution around its mean. A positive (negative) value ofthe skewness implies a distribution with a higher number oflarge (small) values of the parameter than would be expectedfor a Gaussian distribution. The kurtosis measures the rela-tive peakedness or flatness of a distribution compared to anormal distribution, with the same mean and standard devia-tion. A larger kurtosis depicts a more peaked distribution. Fora Gaussian distribution, the skewness and kurtosis are zero,with variances, respectively, equal to (6/N)1/2 and (24/N)1/2,where N is the number of points.

2.4. Wavelet analysis

The wavelet transform (WT) is a newly developed time-frequency analysis method, which has good time resolution athigh frequencies and good frequency resolution at low frequen-cies. It can be viewed as a sort of mathematical microscopebecause different parts of the time series may be examined byautomatically adjusting the focus. In principle, wavelets arethe building blocks of WTs, just as trigonometric functions areused in Fourier transform. In this way, a family of scaled andtranslated wavelets is created, which can catch different fre-quency information contained in a time series. The fact thateach wavelet is a sort of band-pass filter can be used for de-composition of the original time series into different frequencybands, thus allowing a multi-resolution analysis of the signal.

Wavelet could be a powerful tool to analyze multiphase sys-tems which are not only nonlinear but also possess a large

number of frequencies. A number of researchers (Ren et al.,2001; Ellis et al., 2003, 2004) have adopted this analysis to un-derstand the hydrodynamics of fluidized bed. A review of theapplication of the wavelet analysis in different multiphase sys-tems has been given by Drahos et al. (2004). However, it hasrarely been used for liquid–liquid flows. Since liquid–liquidflows exhibit multiscale behavior like any other multiphase sys-tems, leading to multiple components, an appropriate resolu-tion of such multiscale behavior is expected to be helpful tounderstand the complex interfacial distributions.

Accordingly, the Daubechies 4 wavelet analysis has beenused in the present study. The Daubechies wavelets are com-pactly supported wavelets with external phase and highest num-ber of vanishing moments for a given support width. As the sig-nals are decomposed, the frequency range decreases while thescale captured by the signal increases. The details of the calcu-lation have been provided by Takei et al. (2000). For all waveletanalysis, Matlab� with its toolbox has been used. The inher-ent structures of the different flow patterns and their changeswith superficial velocities have been visualized from the de-composed signal.

3. Results and discussions

The experiments are initially conducted at a low velocity ofwater. Keeping the water velocity constant, the kerosene veloc-ity is varied from a low (0.05 m/s) to a high value (1.5 m/s). Theprobe signals are recorded at each combination of phase veloci-ties for a period of 100 s while the interfacial configurations areobserved visually and by photographic means (wherever pos-sible). After the signals are recorded over the entire range ofkerosene velocity, the velocity of water is increased to a highervalue. This is repeated for eight different values of water su-perficial velocity ranging from 0.05 to 1.5 m/s. For each watersuperficial velocity, kerosene superficial velocity is kept con-stant at 10 different values. The results for a few representativeflow rates have been discussed below.

3.1. Flow patterns from probe signals and PDFs

The flow situation as observed visually and from probe sig-nals and the corresponding PDF curves are depicted in a tabularform in Figs. 3–6. In Figs. 3–6 and 8–11 USW and USK denotesuperficial velocities of water and kerosene, respectively. Eachtable represents the flow patterns at different constant values ofwater velocity. The rows of the tables are numbered as 3.1, 3.2,etc. They depict the flow situations for different kerosene ve-locities at the constant value of water superficial velocity. Thedepiction is represented in three parts—(a), (b) and (c) denot-ing the physical appearance of the phase distribution, the probesignals and the corresponding PDF curves, respectively. Themoments of the PDFs are included in the last column of the ta-ble for quantification of the PDFs. In this column the standarddeviation, skewness and kurtosis are denoted by �, S and K,respectively.

At low superficial velocities of water (0.05 m/s) the keroseneis distributed as discrete droplets in the continuous water phase


Fig. No.

USK(m/s)

Visual obs. (a)

Probe signal

(b)

PDF & statistical moments

(c)

3.1 0.05

3.2 0.11

3.3 0.18

3.4 0.25

3.5 0.4

3.6 1.2

0

0.5

1

0

0.5

1

0

0.5

1

0

0.5

1

0

0.5

1

0

0.5

1

40200

40200

40200

402010 30 500

V/ V

max

V/ V

max

V/ V

max

V/ V

max

V/ V

max

V/ V

max

σ 0.0279S -1. 16 K 1.968

σ 0.0788S -1.22 K 1.46

σ 0.0939S 1.91K 3.74

σ 0.0042S 3.68K 26.94

σ 0.0021S 0.24K 26.94

σ 0.0017S -0.32 K 0.267

Bluish

Bluish

Bluish

V/Vmax

V/Vmax

V/Vmax

V/Vmax

V/Vmax

V/Vmax

Blue

Time (s)

Time (s)

Time (s)

Time (s)

40200

0 50 100 150

Time (s)

Time (s)

PDF

PD

F

PD

F

PD

F

PD

F

PD

F

Fig. 3. Visual observations, probe signals and PDFs at USW = 0.05 m/s.

as shown in Fig. 3(3.1a). The droplets are cap shaped or spher-ical and uniformly dispersed in the water medium. The corre-sponding probe signal (Fig. 3(3.1b)) denotes an almost steadyoutput at V/Vmax close to 1 confirming the continuity of thewater phase and a few spikes to lower values of voltage de-picting the occasional bridging of the probes by the kerosenedroplets. The PDF (Fig. 3(3.1c)) is unimodal and skewed to theleft due to the presence of kerosene droplets.

An increase in the kerosene superficial velocity to 0.11 m/sresults in an increase of the size and frequency of droplets in

the flow passage (Fig. 3(3.2a)). The droplets are now mostlyirregular and/or oblate spheroidal and propagate in an er-ratic fashion through water. The corresponding probe signal(Fig. 3(3.2b)) shows an increase in the frequency and the am-plitude of the spikes towards lower voltage values. There is anincrease in the standard deviation of the PDF and a decreasein the peak height in Fig. 3(3.2c) but it still exhibits a negativeskewness. This shows that water remains as the continuousphase while larger and more kerosene droplets contact theprobe due to increased kerosene flow.


Fig. No.

USK(m/s)

Visual obs. (a)

Probe signal

(b)


(c)

4.1 0.05

4.2 0.4

4.3 0.6

4.4 0.9

4.5 1.2

4.6 1.5

Bluish

Bluish

Bluish

Bluish

Bluish

V/Vmax

V/Vmax

σ 0.082 S -0.44 K -0.24

σ 0.067 S 1.21K 1.58

σ 0.021 S 1.64K 10.0

σ 0.005 S -2.96 K 14.9

0

0.5

1

V/V

max

0

0.5

1

V/V

max

V/Vmax

σ 0.024 S -1.68 K 3.6

σ 0.066 S -0.57 K 0.08

402010 30 5000

0.5

1

V/ V

max

0

0.5

1

V/ V

max

0

0.5

1

V/ V

max

0

0.5

1

V/ V

max

V/Vmax

V/Vmax

V/Vmax

Time (s)

40200Time (s)

40200Time (s)

40200Time (s)

40200Time (s)

40200Time (s)

PD

F

PD

F

PD

F

PD

F

PD

F

PD

F


A completely different situation arises on a slight in-crease in kerosene superficial velocity from 0.11 to 0.18 m/s(Fig. 3(3.3)). The flow in this case cannot be discerned visu-ally. The entire passage appears to be blue in color. The changecan best be understood by noting the change in the nature ofthe PDF curve. It is unimodal but the peak has shifted to lowervalues of voltage (V/Vmax < 0.5) and the curve has changedskewness from negative to positive. This probably suggests achange of the continuous phase from water to kerosene due toa rapid coalescence of kerosene chunks but gives no furtherindication about the distribution of the two liquids. Severalresearchers have noted such a changeover during liquid–liquidflow through horizontal pipes (Lovick and Angeli, 2004;

Ioannou et al., 2004, 2005) as well as in stirred vessels(Arashmid and Jeffreys, 1980; Tidhar et al., 1986). They havereported a change from kerosene dispersed in water to waterdispersed in kerosene under certain flow conditions and termedthe phenomena as phase inversion.

On further increase in kerosene velocity the flow passageassumes a uniform bluish appearance and no information canbe extracted visually or by photography. The probe signal(Fig. 3(3.4b)) is almost uniform at V/Vmax close to 0.2 andyields a unimodal PDF with negligible standard deviation buta high positive skewness and large kurtosis (high peakedness).

As the kerosene velocity is increased more, the probe signalappears to be practically a straight line (Figs. 3(3.5–3.6)). The


Fig. No.

USK(m/s)

Visual obs.

(a)

Probe signal

(b)


(c)5.1 0.05

5.2 0.40

5.3 0.90

5.4 1.2

5.5 1.5

σ 0.0021S -0.292 K 0.019

σ 0.0088S -0.477 K 0.756

σ 0.0246S -4.16 K 29.4

σ 0.113 S 0.36K -0.86

σ 0.075 S 2.42K 5.63

Whitish blue

Whitish blue

Whitish blue

Whitish blue

V/Vmax

V/Vmax

V/Vmax

V/Vmax

V/Vmax

0402010 30 500

0.5

1

V/ V

max

0

0.5

1

V/ V

max

0

0.5

1

V/ V

max

0

0.5

1

V/ V

max

0

0.5

1

V/ V

max

Time (s)

402010 30 500Time (s)

40

40

20

20

10 30 500

0

Time (s)

Time (s)

40200Time (s)

PD

F

PD

F

PD

F

PD

F

PD

F


unimodal PDF becomes sharper with lower values of standarddeviation and decreases in height till a certain kerosene super-ficial velocity (0.6 m/s). Subsequently, the skewness shifts toa small negative value but the overall appearance of the PDFremain unchanged.

The signals recorded at other water velocities also show iden-tical trends—a unimodal PDF with negative skewness at lowkerosene velocities, an increasing standard deviation with in-creasing kerosene flow, an abrupt shift of the PDF to V/Vmaxless than 0.5 accompanied by a change in the sign of the skew-ness at a particular kerosene superficial velocity and finally analmost straight line PDF with minimum spread at lower valuesof V/Vmax at higher values of kerosene velocities. Although theresults have been noted for several water velocities, only oneadditional series have been shown in Fig. 4 to avoid repetition.

The situation alters slightly at higher water superficial ve-locities (Figs. 5 and 6 at water superficial velocities of 0.9 and

1.2 m/s, respectively). The kerosene appears to be dispersed asfine droplets unlike the irregular shapes at lower water super-ficial velocity. The probe signals are almost straight lines atV/Vmax close to unity and the PDFs are unimodal with negli-gible spread due to the short residence time of the tiny droplets.The spread increases with increasing kerosene superficial ve-locity probably due to the higher area occupied by kerosene.The flow passage appears to be white/whitish blue under theseconditions. In Fig. 5, the shift of the PDF curve from nega-tive to positive skewness occurs when kerosene superficial ve-locity is increased from 0.9 to 1.2 m/s (Fig. 5(5.3–5.4)). It isaccompanied by a large increase in the value of standard de-viation probably due to the comparable holdups of the twophases under these conditions. At higher kerosene velocities, thepeak of the PDF at V/Vmax < 0.5 indicates kerosene predomi-nance in the flow passage while the value of standard deviationshow significant presence of the water phase as well. At water


Fig. No.

USK(m/s)

Visual obs.

(a)

Probe signal

(b)


(c)

6.1 0.05

6.2 0.4

6.3 1.5

V/Vmax

σ 0.0017S – 0.65 K 0.90

V/Vmax

σ 0.0069S – 0.33 K 0.19

V/Vmax

σ 0.0364S –1.36 K 3.17

Whitish

0

0.5

1

V/ V

max

0

0.5

1

V/ V

max

0

0.5

1

V/ V

max

40200Time (s)

40200Time (s)

40200Time (s)

PD

F

PD

F

PD

F


superficial velocity 1.2 m/s (Fig. 6), the shift of skewness asnoted earlier is not observed in the range of flow rates studies.The spread of the PDF increases but its position does not shiftmuch thus indicating that water remains as the continuous phasein the range of the phase velocities studied.

The aforementioned figures show that the visual appearanceas well as the probe signals prove to be less effective at higherphase velocities. The appearance and moments of the PDFssuggest the probable distribution with regard to the continuousphase (peak position) and the extent of the distribution of thesecond phase (standard deviation and skewness). The shift inthe position of PDF and sign of skewness beyond a particularkerosene velocity marks the transition from dispersed flow. Thisoccurs at progressively higher kerosene velocities as the watervelocity is increased. However, the PDF gives no idea aboutthe nature of distribution of the two phases at high kerosenevelocities. From the survey of the past studies, it was expectedthat beyond phase inversion the water would be distributed asdroplets in the continuous kerosene phase. The PDFs do notrule out this possibility but give no definite proof of such adistribution. The nature of PDF obtained is indicative of eithera fine dispersion of water droplets or separated flow with awater film at the wall.

3.2. Wavelet analysis of probe signal

The wavelet analysis has next been performed to supple-ment the information obtained from PDFs under different flow

conditions. Since the analysis can decompose a signal into dif-ferent frequency bands with different time resolutions, they areexpected to yield additional information about the interfacialconfigurations. As mentioned earlier, the Daubechies 4 waveletanalysis has been used to decompose the normalized probe sig-nals into 5 levels, each with a different frequency band. Onefull-decomposed signal with five details and one approximationis depicted in Fig. 7. In the figure d1 reflects the smallest scaleand the highest frequency band; d2, d3, etc. represent progres-sively lower frequency bands and the approximation, a5 reflectsthe large-scale resolution.

In our study, we have primarily concentrated our attention ond1 and a5 in order to understand the fluctuations due to passingdroplets and wavy interfaces, respectively. It is expected thatthe high frequency fluctuations will be caused by passage ofdroplets while the large scale low frequency fluctuations willarise from any waviness of the continuous interface. Accord-ingly the d1 and a5 components of the normalized probe signalsreported in Figs. 3–6 are presented in Figs. 8–11, respectively.

At a water superficial velocity of 0.05 m/s, there are largefluctuations for low kerosene velocities at detail d1 while theapproximation a5 remains relatively smooth. This indicates thepassage of kerosene drops through the water medium. The tran-sition from dispersed flow (abrupt change of the sign of skew-ness and peak position in the PDF curve of Fig. 3(3.3c) at akerosene superficial velocity of 0.18 m/s) is marked by a sud-den rise in the fluctuations in a5 while those of d1 are relativelyunchanged (Fig. 8(8.3)). This is an indication of a change inflow pattern from dispersed to separated flow with an extremely


Fig. 7. Wavelet analysis of a typical signal obtained from conductivity probe.

wavy interface. Moreover, the fluctuations at both the levelsshow some intermittent character with nearly periodic peaksin a5. This implies the occurrence of large chunks of kerosenedrops, which share a deformable interface with water similarto the slug/churn transition in gas–liquid cases. Interestingly,a further increase in kerosene velocity is marked by a suddenreduction in the fluctuations of both d1 and a5 for all highervalues of kerosene superficial velocity. This indicates a smoothinterface between the two liquids and does not support the ex-istence of inverted dispersed flow (water dispersed in kerosene)under these conditions. At a water superficial velocity of 0.3 m/s(Fig. 9) the d1 fluctuation increases with the increase inkerosene superficial velocity while the a5 fluctuations remainlow in the entire range of dispersed flow pattern. This de-notes an increase in the density of kerosene drops and showsthat the increasing standard deviation in the PDF occurreddue to the dispersed kerosene in water. The slight increase influctuation of a5 with increasing kerosene superficial velocitymay be due to continuous change in kerosene–water interface.The termination of dispersed flow as noted from the PDF(Fig. 4(4.4c)) is marked by an increased waviness in a5 whilethe fluctuation of d1 appear to be less dense (Fig. 9(9.4)). Thisagain indicates that the deformed kerosene droplets probablycoalesce to form a continuous kerosene core separated by awater film. Subsequently the fluctuations in both a5 and d1

decrease drastically in agreement to the observations at lowervalues of water velocity.

There are negligible fluctuations in both the detail and ap-proximation in the dispersed flow pattern at water superficialvelocity 0.9 m/s (Fig. 10) and 1.2 m/s (Fig. 11). This is in agree-ment with the almost straight line PDF reported in Figs. 5 and6. The transition from dispersed flow in Fig. 10 is marked byan increase in the fluctuations of both d1 and a5 (Fig. 10(10.4)).While the a5 fluctuations remain at higher kerosene velocities,the d1 fluctuations decrease. This denotes a flow pattern withwavy interface rather than dispersed droplets. At water super-ficial velocity 1.2 m/s, the fluctuations in the detail d1 progres-sively increases with kerosene velocity while the approxima-tion is relatively smooth thus confirming the contribution ofincreased droplet size and frequency in the increased spread ofthe PDF.

It may be noted that the wavelet analysis has confirmedthe existence of dispersed flow at low kerosene velocities butrules out inverted dispersed pattern at high flow rates. Onthe contrary, it seems to indicate the existence of a separatedflow pattern namely, annular flow under such conditions. Italso denotes the existence of a smoother interface at higherkerosene and low water velocities. However, the definite ex-istence of annular flow under such conditions needs furtherinvestigations.


Fig. No.

USK(m/s)

d a1 5

8.1 0.05

8.2 0.11

8.3 0.18

8.4 0.25

8.5 0.4

8.6 1.2

0.5

0.75

1

0.5

0.75

1

0.2

0.45

0.7

-0.05

0

0.05

-0.05

0

0.05

-0.05

0

0.05

-0.05

0

0.05

-0.05

0

0.05

-0.05

0

0.05

0

0.25

0.5

0

0.25

0.5

0 500 1000

0 500 1000

0 500 1000

0 500 10000 500 1000

0 500 1000

0 500 1000

0 500 1000

0 500 1000

0 500 1000

0 500 1000

0

0.25

0.5

0 500 1000

Fig. 8. d1 and a5 from wavelet analysis from probe signal at USW = 0.05 m/s.

3.3. Wall mounted conductivity probe

From the probe signals and their PDF and wavelet analysisat high kerosene and low water velocities, it is postulated thatan annular distribution of the two liquids with water film at thewall and kerosene at the core exists at this condition. However,the signal from the parallel wire probe gives information aboutthe phase distribution “averaged” over the diametral plane. It isdifficult to make an idea regarding the radial phase distributionfrom this probe.

In order to verify the presence of a continuous water film atthe wall, a system comprising of wall mounted point electrode

probes has been used in the study. A total of eight stainlesssteel wires of 1.2 mm diameter have been flush mounted onthe inside pipe wall at uniform distances in the same horizon-tal plane as illustrated in Fig. 12. The probe tips are markedserially from 1 to 8. Initially the voltage has been measuredbetween different combinations of probe tips during the flowof single phase water and kerosene, respectively. It has beennoted that during single phase flow the voltage difference doesnot depend on velocity of the corresponding phases. The aver-age voltage varies between 6.49 and 6.54V for only water flowand between 4.4 and 4.66V for only kerosene flow throughthe pipe.


Fig. No.

USK(m/s)

d a1 5

9.1 0.05

9.2 0.4

9.3 0.6

9.4 0.9

9.5 1.2

9.6 1.5

0.5

0.75

1

-0.15

0.00

0.15

-0.15

0.00

0.15

-0.15

0.00

0.15

-0.15

0.00

0.15

-0.15

0.00

0.15

-0.15

0.00

0.15

0.5

0.75

1

0 500 1000

0 500 1000

0 500 1000

0 500 1000

0 500 1000

0 500 10000 500 1000

0 500 1000

0 500 1000

0 500 1000

0 500 1000

0 500 1000

0.5

0.75

1

0.30

0.55

0.80

0.30

0.55

0.80

0.30

0.55

0.80


The experiments have next been repeated for the simultane-ous flow of the two liquids. The measurements are confinedto high flow rates beyond the dispersed flow pattern. The datafor one pair of phase velocities (water superficial velocity of0.3 m/s and kerosene superficial velocity of 1.2 m/s) is reportedin Table 2 where USW and USK denote the superficial veloci-ties of water and kerosene, respectively. It shows that the volt-age data for two adjacent tips is almost constant (5.96V) for allprobe pairs (1–2, 2–3, . . . , 7–8) and close to that obtained foronly water. On the other hand, the voltage between nonadjacentpairs of probe is much lower and the least value is obtained foropposite pairs of probe (2–6, 1–5, 3–7 and 4–8). This confirmsthe presence of a continuous water film at the wall.

In order to verify the composition of the core, an addi-tional point electrode probe made of the same wire as the wallmounted probes has been incorporated in the measurement sys-tem. It is inserted in the pipe 6 cm above the wall-mountedprobe as shown in Fig. 13. The probe is bent by 90◦ towardsthe bottom to ensure that the tip is located at the center of thepipe in the same horizontal plane as the flush mounted probes.The whole length of the wire has been insulated by a coat ofvarnish and then by an epoxy coating except the tip. The tipdiameter is 0.1 mm and this is marked as “0” in Fig. 13. Thevoltage difference between different wall mounted probes andthe central probe (0) has next been noted for the same combina-tion of phase velocities. The result for one representative case


Fig. No.

USK(m/s)

d a1 5

10.1 0.05

10.2 0.4

10.3 0.9

10.4 1.2

10.5 1.5

-0.20

0.00

0.20

-0.20

0.00

0.20

-0.20

0.00

0.20

-0.20

0.00

0.20

0 500 1000

0 500 1000

0 500 1000

0 500 1000

0.5

0.75

1

0.5

0.75

1

0.5

0.75

1

0 500 1000

0 500 10000.3

0.55

0.8

0.3

0.55

0.8

-0.20

0.00

0.20

0 500 1000 0 500 1000

0 500 1000

0 500 1000


is depicted in Table 3. It shows a drastic reduction in voltagebetween 0 and X (X = 0–8) as compared to that between thetwo consecutive wall mounted probe tips. This confirms that awater film exists at the pipe wall and kerosene flows throughthe central core. Since kerosene has a lower conductivity ascompared to water, the voltage is much lower for 0–X and thevoltage between the two opposite point tips like 1–5, 3–7, etc.,is slightly greater than 0–X.

Similar results have been obtained over the entire range ofphase velocities where kerosene appears to be the continuousphase. From the above observations, it has been confirmed thatthe flow is annular at high kerosene velocities. The water existsas a film while kerosene flows through the core.

3.4. Flow pattern map

Further attempts have been made to represent the flow pat-terns observed under different conditions in the form of a flowpattern map. The probe signals have shown that at low phasevelocities, kerosene flows as discrete droplets in the continuouswater medium. The droplets are cap shaped or oblate spheroidaland resemble the bubbly flow pattern in gas–liquid flows. Itis, therefore, named as the “bubbly flow pattern”. The dropletsdecrease in size and become finely dispersed at higher watersuperficial velocities (close 1 m/s) due to the shearing force cre-ated by the increased turbulence of the water phase. This regionis characterized by almost straight line PDFs at high voltage


Fig. No.

USK(m/s)

d1 a5

11.1 0.05

11.2 0.4

11.3 1.5

0.5

0.75

1

0.5

0.75

1

0 500 1000

0.5

0.75

1

0 200 400 600 800 1000

-0.08

0

0.08

-0.08

0

0.08

-0.08

0

0.08

0 500 1000 0 500 1000

0 200 400 600 800 1000

0 200 400 600 800 1000


3 7

1

5

4

2 8

6

Pipe wall

Wall mountedprobes

Fig. 12. Wall mounted conductivity probe.

Table 2Measurements from wall mounted probe

Probe Water (USW =0.3 m/s)

Kerosene (USK =0.4 m/s)

Water–kerosene(USW = 0.3,USK = 1.2 m/s)

Average volt Average volt Average volt

1–2 6.53 4.66 5.961–3 6.50 4.60 5.801–4 6.49 4.52 5.501–5 6.49 4.59 5.032–3 6.53 4.66 5.962–4 6.50 4.61 5.802–5 6.49 4.40 5.502–6 6.49 4.59 5.033–4 6.54 4.66 5.963–5 6.50 4.60 5.803–6 6.49 4.40 5.503–7 6.49 4.59 5.03

Central probe

Pipe wall

Wall mounted probe

Fig. 13. Centerline point and wall mounted conductivity probe.

Table 3Measurements from centerline and wall mounted probes

Probe Average volt

0–1 4.710–2 4.700–3 4.700–4 4.710–5 4.710–6 4.700–7 4.710–8 4.71

values and negligible fluctuations in the d1 and a5 curves ofthe wavelet analysis. It is termed as the “dispersed bubbly flowpattern”.

At high kerosene velocities, the rapid coalescence of the oilchunks form a continuous core which pushes the water to flowas an annular film along the walls. This leads to the annular flowpattern as noted from the different designs of the conductivity


0.01

0.1

1

10

0.01 0.1 1 10

Superficial velocity of kerosene, m /s

Su

per

fici

al v

elo

city

of

wat

er, m

/s

Dispersed Bubbly

Bubbly

Churn Turbulent

Core Annular

Fig. 14. Flow regime map.

probe. Researchers have named this pattern as “core annular”flow in literature.

The transition between bubbly and core annular flow regimesappears to occur through a phase distribution where water isthe continuous medium and is gradually pushed to the wallsdue to the coalescing kerosene droplets. It comprises of irreg-ular chunks of kerosene and water with continuously changinginterface between them and marks the changeover from wa-ter dominating to kerosene dominating flow distribution. Bothd1 and a5 exhibit random fluctuations, which shows that thepattern is chaotic and irregular like the churn flow pattern ofgas–liquid flows. This distribution is therefore named as the“churn turbulent” flow pattern. It exists over a wider range athigher phase velocities.

It is quite interesting to note that at lower superficial velocity(0.05 m/s) of water the asymptote for the boundary betweenchurn turbulent and core annular flow lies at a value of kerosenesuperficial velocity of about 0.25 m/s. This corresponds to avalue of unity of the dimensionless kerosene velocity wherethe kerosene velocity has normalized with

√�K/(�W − �K) as

suggested by Wallis (1969). In the above expression �K and �Wcorresponds to the densities of kerosene and water, respectively.This is in agreement to the onset of the annular flow pattern ingas–liquid flows as suggested by Wallis (1969).

The information thus obtained from the conductivity probesis represented graphically in the form of a flow pattern map(Fig. 14). The superficial velocity of kerosene and water hasbeen selected as the axes of the map in agreement to the conven-tion adopted for gas–liquid flows. It may be noted that the flowpatterns were studied in an acrylic resin pipe and the ranges ofexistence of the different patterns might be influenced by thepipe material. However, this would not alter the gross charac-teristics of flow.

4. Conclusions

The flow patterns during two-phase kerosene–waterflow through a vertical conduit has been identified using

the conductivity probe method. The normalized time series dataof parallel wire probe has been analyzed by PDF and waveletanalysis. The quantification of the PDFs has been accomplishedby the statistical moments namely, standard deviation, skewnessand kurtosis. The analysis has showed that at low flow rates ofkerosene, kerosene flows as droplets in the continuous waterphase. In convention to gas–liquid flow, this flow pattern hasbeen named as the bubbly flow pattern. At high flow rates ofkerosene, the analysis shows that there may exist a separatedflow pattern like core annular flow. The continuity of the waterlayer at the pipe wall was examined by applying wall mountedpoint probes. To examine the existence of kerosene in the centerof the pipe at high flow rates of kerosene, a point probe has alsobeen inserted in the center of the pipe. The probe measurementshave suggested that the flow pattern is core annular with wateras a thin layer or film at the pipe wall and kerosene in thecenter of the pipe. In the gas–liquid flow, the bubbly to annularflow pattern transition proceeds via slug-churn flow regimes. Inthe present study no such clear slug has been observed. Herethe transition has taken place via a flow pattern consisting ofirregular shaped chunks and bubbles of kerosene in water. Thisflow pattern has been named as the churn turbulent flow pattern.Interestingly the transition from churn-turbulent to the annularflow pattern is in agreement to the onset of annular flow ingas–liquid cases.

References

Al-Deen, M.F.N., Hamad, F.A., Bruun, H.H., 1998. Kerosene/water two-phaseflow in vertical and inclined pipes. Proceedings of the Third InternationalConference on Multi-phase Flow, Lyon, 1998.

Andreussi, P., Bendiksen, K., 1989. An investigation of void fraction in liquidslugs for horizontal and inclined gas liquid pipe flow. International Journalof Multiphase Flow 15, 937–946.

Angeli, P., Hewitt, G.F., 1998. Pressure gradient in liquid–liquid flows.International Journal of Multiphase Flow 24, 183–1203.

Angeli, P., Hewitt, G.F., 2000. Flow structure in horizontal oil–water flow.International Journal of Multiphase Flow 26, 1117–1140.

Arashmid, M., Jeffreys, G.V., 1980. Analysis of the phase inversioncharacteristics of liquid–liquid dispersions. A.I.Ch.E. Journal 26, 51–55.

Arirachakaran, S., Oglesby, K.D., Malinowsky, M.S., Shoham, O., Brill, J.P.,1989. An analysis of oil/water flow phenomena in horizontal pipes. In: SPEProduction Operating Symposium, SPE Paper 18836, Oklahoma, March13–14, pp. 155–167.

Brauner, N., 2001. The prediction of dispersed flows boundaries inliquid–liquid and gas–liquid systems. International Journal of MultiphaseFlow 27, 885–910.

Brauner, N., Maron, M.D., 1992. Flow pattern transitions in two-phaseliquid–liquid flow in horizontal tubes. International Journal of MultiphaseFlow 18, 123–140.

Brauner, N., Ullmann, A., 2002. Modelling of phase inversion phenomenonin two-phase pipe flow. International Journal of Multiphase Flow 28,1177–1204.

Brown, R.A.S., Govier, G.W., 1961. High-speed photography in the study oftwo-phase flow. Canadian Journal of Chemical Engineering 39, 159–164.

Charles, M.E., Govier, G.W., Hodgson, G.W., 1961. The horizontal pipelineflow of equal density oil–water mixture. Canadian Journal of ChemicalEngineering 39, 27–36.

Cheng, H., Hills, J.H., Azzopardi, B.J., 2002. Effects of initial bubble sizeon flow pattern transition in a 28.9 mm diameter column. InternationalJournal of Multiphase Flow 28, 1047–1062.


Das, G., Das, P.K., Purohit, N.K., Mitra, A.K., 2000. Phase distributionof gas–liquid mixtures in concentric annuli-inception and termination ofasymmetry. International Journal of Multiphase Flow 26, 857–876.

Drahos, J., Ruzicka, Marek, C., 2004. Problems of time series analysisin characterization of multiphase flows. Fifth International Conferenceon Multiphase Flow, ICMF’04, Paper no. K04, Yokohama, Japan, May30–June 4.

Ellis, N., Briens, L.A., Grace, J.R., Bi, H.T., Lim, C.J., 2003. Characterizationof dynamic behaviour in gas–solid turbulent fluidized bed using chaos andwavelet analyses. Chemical Engineering Journal 96, 105–116.

Ellis, N., Bi, H.T., Lim, C.J., Grace, J.R., 2004. Influence of probe scale andanalysis method on measured hydrodynamic properties of gas-fluidizedbeds. Chemical Engineering Science 59, 1841–1851.

Farrar, B., Bruun, H.H., 1989. Interaction effects between a cylindrical hot-film anemometer probe and bubbles in air/water and oil/water flow. Journalof Physics E: Scientific Instruments 22, 114–123.

Farrar, B., Bruun, H.H., 1996. A computer based hot-film technique used forflow measurements in a vertical kerosene–water pipe flow. InternationalJournal of Multiphase Flow 22, 733–751.

Flores, J.G., Sarica, C., Chen, T.X., Brill, J.P., 1998. Investigation of holdupand pressure drop behaviour for oil–water flow in vertical and deviatedwells. Transactions of ASME 120, 8–14.

Govier, G.W., Sullivan, G.A., Wood, R.K., 1961. The upward vertical flow ofoil–water mixtures. Canadian Journal of Chemical Engineering 9, 67–75.

Guzhov, A., Grishin, A.D., Medredev, V.F., Medredeva, O.P., 1973. Emulsionformation during the flow of two liquids in a pipe. Neftyanoe Khozyaistvo8, 58–61 (in Russian).

Hamad, F.A., Brunn, H.H., 2000. Evaluation of bubble/drop velocity by asingle normal hot-film placed in a two-phase flow. Measurement Scienceand Technology 11, 11–19.

Hamad, F.A., Imberton, F., Brunn, H.H., 1997. An optical probe formeasurement in liquid–liquid two-phase flow. Measurement Science andTechnology 8, 1122–1132.

Hamad, F.A., Pierscionek, B.K., Brunn, H.H., 2000. A dual optical probe forvolume fraction, drop velocity and drop size measurements in liquid–liquidtwo-phase flow. Measurement Science and Technology 11, 1307–1318.

Hasson, D., Mann, U., Nir, A., 1970. Annular flow of two immiscible liquids.I. Mechanisms. Canadian Journal of Chemical Engineering 48, 514–520.

Hasson, D., Orell, A., Fink, M., 1974. A study of vertical annular liquid–liquidflow—Part I, laminar conditions. Multi-Phase Flow Systems Symposium,The Institution of Chemical Engineers Symposium Series no 38, Paperno. A5, Glasgow, pp. 1–16.

Ioannou, K., Hu, B., Matar, Omar K., Hewitt, G.F., Angeli, P., 2004.Phase inversion in dispersed liquid–liquid pipe flows. Fifth InternationalConference on Multiphase Flow, ICMF’04, Paper no. 108, Yokohama,Japan, May 30–June 4.

Ioannou, K., Nydal, Ole, J., Angeli, P., 2005. Phase inversion in dispersedliquid–liquid flows. Experimental Thermal and Fluid Science 29, 331–339.

Jin, N.D., Nie, X.B., Ren, Y.Y., Liu, X.B., 2003. Characterization of oil/watertwo-phase flow patterns based on nonlinear time series analysis. FlowMeasurement and Instrumentation 14, 169–175.

Jones, O.C., Delhaye, J.M., 1976. Transient and statistical measurementtechniques for two phase flows: a critical review. International Journal ofMultiphase Flow 3, 89–116.

Jones, O.C., Zuber, N., 1975. The interrelation between void fractionfluctuations and flow pattern in two phase flow. International Journal ofMultiphase Flow 2, 273–306.

Li, H., 2002. Application of wavelet multi-resolution analysis to pressurefluctuations of gas–solid two-phase flow in horizontal pipe. PowderTechnology 125, 61–73.

Liu, T.J., Bankoff, S.G., 1993. Structure of air–water bubbly flow in a verticalpipe: I. Liquid mean velocity and turbulence measurements. InternationalJournal of Heat and Mass Transfer 36, 1049–1060.

Lovick, J., Angeli, P., 2004. Experimental studies on the dual continuousflow pattern in oil–water flows. International Journal of Multiphase Flow30, 139–157.

Miya, M., Woodmansee, D.E., Hanratty, T.J., 1971. A model of roll wavesin gas liquid flow. Chemical Engineering Science 26, 1915–1931.

Mukherjee, H., Brill, J.P., Beggs, H.D., 1981. Experimental study of oil–waterflow in inclined pipes. Transactions of the ASME 103, 56–66.

Nadler, M., Mewes, D., 1995. The effect of gas injection on the flow of twoimmiscible liquids in horizontal pipes. Chemical Engineering Technology18, 156–165.

Pacek, A.W., Moore, I.P.T., Nienow, A.W., 1994. Video technique formeasuring dynamics of liquid–liquid dispersion during phase inversion.A.I.Ch.E. Journal 40, 1940–1949.

Ren, J., Mao, Q., Li, J., Lin, W., 2001. Wavelet analysis of dynamic behaviourin fluidized beds. Chemical Engineering Science 56, 981–988.

Rensen, J., Luther, S., Vries de, J., Lohse, D., 2005. Hot film anemometry inbubbly flow I: bubble-probe interaction. International Journal of MultiphaseFlow 31, 285–301.

Russell, T.W.F., Hodgson, G.W., Govier, G.W., 1959. Horizontal pipeline flowof mixtures of oil and water. Canadian Journal of Chemical Engineering37, 9–17.

Serizawa, A., Kataoka, I., Michiyoshi, I., 1975. Turbulence structure ofair–water bubbly flow—II. Local properties. International Journal ofMultiphase Flow 2, 235–246.

Simmons, M.J.H., Azzopardi, B.J., 2001. Drop size distributions in dispersedliquid–liquid pipe flow. International Journal of Multiphase Flow 27,843–859.

Sun, X., Kim, S., Cheng, L., Ishii, M., Beus, S.G., 2004. Interfacial structuresin confined cap-turbulent and churn-turbulent flows. International Journalof Heat and Fluid Flow 25, 44–57.

Takei, M., Ochi, M., Horii, K., Li, H., Saito, Y., 2000. Discrete waveletsauto-correlation of axial turbulence velocity in spiral single phase flow.Powder Technology 112, 289–298.

Tidhar, M., Merchuk, J.C., Sembira, A.N., Wolf, D., 1986. Characteristics of amotionless mixer for dispersion of immiscible fluids—II. Phase inversionof liquid–liquid systems. Chemical Engineering Science 41, 457–462.

Trallero, J.L., 1995. Oil–water flow patterns in horizontal pipes. Ph.D. Thesis,The University of Tulsa.

Valle, A., Kvandal, H., 1995. Pressure drop and dispersion characteristicsof separated oil–water flow. In: Proceedings of the First InternationalSymposium on Two-Phase Flow Modelling Experimentation, Rome, Italy,9–11 October, pp. 583–591.

Vigneaux, P., Chenais, P., Hulin, J.P., 1988. Liquid–liquid flows in an inclinedpipe. A.I.Ch.E. Journal 34, 781–789.

Vince, M.A., Lahey Jr., R.T., 1982. On the development of an objective flowregime indicator. International Journal of Multiphase Flow 8, 93–124.

Wallis, G.B., 1969. One-dimensional Two-phase Flow. McGraw-Hill, NewYork.

Wang, S.K., Lee, S.J., Jones, O.C., Lahey, R.T., 1987. 3-D turbulence structurephase distribution measurements in bubbly two-phase flows. InternationalJournal of Multiphase Flow 13, 327–343.

Flow regime identification of two-phase liquid--liquid upflowthrough vertical pipeIntroductionExperimental setup and procedureMeasurement techniqueMethod of analysisPDF analysisWavelet analysis

Results and discussionsFlow patterns from probe signals and PDFsWavelet analysis of probe signalWall mounted conductivity probeFlow pattern map

ConclusionsReferences

flowregimeidentiﬁcationoftwo-phaseliquid–liquidupﬂow ......flow of a mixture of two immiscible...

Documents