Berto La 2003

Two-Phase Flow Measurement Techniques
Volfango Bertola
Laboratoire de Physique Statistique, Ecole Normale Superieure, Paris, France
Abstract. In these notes the most common measurement techniques for two-phase flows
are reviewed. The working principles and the configurations of instruments for void frac-
tion measurements, flow visualization and velocity measurements are presented; in detail:
radiation attenuation, optical and electrical impedance techniques for void fraction meas-
urement; tomographic and time-average visualization techniques; velocity measurements
from signal cross-correlation, hot film anemometry, particle image velocimetry.
1 Introduction
Two-phase flow is one of the most common flows in nature as well as in industrial applications; it
covers gas-solid, liquid-liquid, solid-liquid and gas-liquid flows. Among these, gas-liquid flow,
which also includes the whole subject of boiling and condensation, is probably the most relevant
topic, and can be encountered in a wide range of industrial applications including evaporators,
boilers, distillation towers, chemical reactors, condensers, oil pipelines, nuclear reactors, etc.
Thus, the early research on two-phase flow was carried out on gas-liquid flow; anyway, the theo-
retical framework developed for gas-liquid flow can be often extended to other two-phase
systems, such as liquid-liquid flow.
Advancements in the knowledge of two-phase flow need a close working relationship among
experimentalists, theoreticians and numerical analysts. In particular, experiments are necessary to
verify the reliability and the limitations of both mathematical models and numerical codes; fur-
thermore, they allow capturing the physics of phenomena, and therefore provide a basis for their
theoretical description.
Measurement techniques in two-phase flows are quite different from those of single-phase
flows: in fact, there exist peculiar quantities ofthis kind of flows, such as the void fraction and the
interfacial area concentration, which require a specifically conceived instrumentation. On the
other hand, measurement techniques which have been developed for single-phase flow measure-
ments, such as hot wire anemometry or particle image velocimetry, generally cannot be used in
two-phase flows as they are, but require some modifications.
2 Void Fraction Measurements
Void fraction is a dimensionless quantity indicating the fraction of a geometric or temporal do-
main occupied by the gas phase, and it is probably the most significant quantity one can measure
in two-phase flow. Although the first void fraction measurements are dated back to the 1940s,
they still play the most important role in today's experiments: a recent statistical study of the
experimental research on two-phase flow shows that almost any journal article on this subject
V. Bertola (ed.), Modelling and Experimentation in Two-Phase Flow

© Springer-Verlag Wien 2003
282 V. Bertola
presents void fraction data (Chanson, 2002). The importance is even greater in applications,
where the void fraction is often a key parameter in design as well as in fmancial issues Gust think
of the petroleum industry, where one of the main problems is the oil-water-gas flow metering).
On the other hand, void fraction measurements are essential to validate the predictions of mathe-
matical models.
According to the different defmitions of void fraction (related to the choice of the spatial or
temporal domain), different instruments have been developed; in particular, one can distinguish:
- local instruments, which directly measure the phase density function or the local, time-
average void fraction;
- integral instruments, which measure space-averaged void fractions.
Void fraction measurement techniques are based on various principles: usually, instruments
are sensitive to some physical property which is different for the two phases, such as the fluid
density or the electrical conductivity.
2.1 Void Fraction Definition

The different definitions of void fraction can be placed in a common mathematical framework
introducing the so-called phase density function, which is the characteristic function of the geo-
metrical subsets of the controlled volume occupied by one of the two phases. Consider a three
dimensional geometrical domain occupied by a two-phase mixture (e.g. gas-liquid) at time t; the
set D of the internal points of such domain, identified by their position x, can be divided into two
subsets D 0 and DL, occupied by the gas phase and by the liquid phase, respectively. The phase
density function Pk (k = G, L) is the characteristic function of these subsets:
(1)
By averaging the gas density function over different spatial or temporal domains one can ob-
tain four defmitions for the void fraction.
Local void fraction. The local void fraction in a given point x of space represents the ratio be-
tween the time during which the point is occupied by the gas and the total duration of the
observation T; from Eq. (1), it is the time average overT of the gas density function P0 (x, t):
ca(x,T)=_!_
T
r Pa(x,t)dt
Jr
(2)
Since two-phase flow is intrinsically inhomogeneous, the value of Eo a priori is a function of

the integration time T; yet, in case of steady flow conditions, Eo can be considered constant with
respect to T if T is large enough, so that:
sa{x)=_!_ r Pa(x,t)dt (3)

T Jr
Figure 1 shows an example of the local void fraction fluctuations with respect to the integra-
tion time for a steady air-water flow in a horizontal pipe.
Two-Phase Flow Measurement Techniques 283
OJ
T [s]
Figure I. Local void fraction nu tuation s. int gration time.
One-dimensional void fraction. The one-dimensional void fraction comes from averaging the
phase density function along a straight line at a given time:
&c 1(t)=_!_ f Pc(x,t)dl= Lc(t) (4)

' L JL L
Thus, it represents the ratio between the length occupied by the gas and the total length L.
Very often, the value of Eo,t depends both on the length, and on the orientation of the line: there-
fore, the local flow structure should be always taken into account in order to avoid
misunderstanding.
Two-dimensional void fraction. The two-dimensional void fraction, or cross-sectional average

void fraction, comes from averaging the phase density function over a given area A at a given
time:
&c 2 (t)= _!_

' A
JPc(x,t)da = Ac(t)
A A
(5)
It is the ratio between the area occupied by the gas and the total area A at time t. For the flow
in pipes (or rectangular channels), a natural choice is the cross-sectional area perpendicular to the
pipe axis.
Three-dimensional void fraction. The three-dimensional void fraction, or average void fraction
over a control volume, is the ratio between the volume of the gas and the whole control volume,
and is defmed as:
284 V. Bertola
(6)
Double-averaged void fractions. Space and time averages can be combined to give double-
averaged void fractions. For a well-known property of double integrals of measurable functions,
space and time average are commutative: in other words, changing the sequence of the operations
does not change the result (Ishii, 1975; Delhaye and Achard, 1976). Therefore, one can defme
three double-averaged void fractions:
&1 =~ IT{~ 1Pa(x,t)dl} dt = ~ I{~ IT Pa(x,t)dt} dl = L; (7)
&2 =_!_ f {_!_J Pa(x,t)da}dt=_!_J {_!_ f Pa(x,t)dt}da= Aa (8)

TJTAA AATJT A
&3 =_!_ f {_!_ f Pa(x,t)dv}dt=_!_ f {_!_ f Pa(x,t)dt}dv= Va (9)

T JT V Jv V Jv T JT V
Unless otherwise declared, the quantity which is usually referred to as "void fraction" in the
literature is the two dimensional (cross-sectional) and time average void fraction, and is generally
denoted by the symbols E or, less frequently, a.
2.2 Radiation Attenuation Techniques

These methods are based on the interaction between condensed matter (in our case, fluids) and
either radiation or particles. Since particles (or, alternatively, the radiation wavelength) are very
small as compared with atoms, interaction with matter occurs only if they hit or at least pass very
close to an atom; as a result, the particle may deviate, bounce, or hit and eventually destroy the
atom. For each fundamental interaction (strong, electro-weak and gravitational) one can define
an area surrounding the atomic nucleus where interactions occur (Krane, 1988), which is known
as cross-section, and is measured in barn (1 barn= 10·28 m2).
For simplicity, let's consider a uniform, steady-state flux <I> of particles entering a volume
Adx perpendicular to the surface A, so that N =<I>A is the number of entering particles per unit
time; the number of interactions after a length dx (that is, the beam attenuation) is:
dN = -<I>Ana dx (10)
where n is the number of nuclei per unit volume and cr the cross-section of the interaction. This
leads to the well known exponential decay law of a monochromatic radiation:
(11)
where A. is the so-called interaction length, which is characteristic of each material, and can be
easily expressed as a function of the density p, of the atomic mass MA and of Avogadro's num-
berNAv:
source
w
.........
detector
Figure 2. Measurement of nc-dimcn ional void fraction by radiation attenuation.
I MA
A.=-=--'-'- (12)
an apNAv
Since the interaction length is different for different fluids, a radiation beam can be used to
measure the one dimensional void fraction in two-phase flow (either instantaneous or time-
average), as shown in figure 2. In this case, there are three contributions to the beam attenuation,
those of the two fluids and that of the pipe walls:
N = N 0 exp(- ~: }xp(- (1 ~: ~ }xp(- ~:) (13)
where w is the wall thickness and d the pipe diameter. This kind of instrument generally uses a y-
ray source, and therefore is known as y-densitometer.
For gas-liquid flow at relatively low pressure, Eq. (13) simplifies because the last exponential
can be neglected. Let No and NL indicate respectively the number of emerging particles one can
measure in case of gas flow (E =I) and in case ofliquid flow (E = 0):
N G = N 0 exp(- ~: }xp(~ ~ )
(14)
N L = N 0 exp(- ~:}xp(- ~ J
Then, the void fraction can be calculated after one measurement in two-phase flow as follows:
(15)
286 V. Bertola
s ource
_,\)
linear~~;:
sou rce
:;=~
----
----
,-""\·
\.---'"'
.... , , ,,
.... ~
.... -.,.
- ----
-- -- - -- <:::>' ,'
A 8 (j P.M. tubes
Figure 3. Configurations of radiation sources for two-dimensional void fraction measurements.
By replacing the beam with a linear source (a radiation sheet, as shown in figure 3a), one can
measure the cross sectional average (two-dimensional) void fraction. Unfortunately, in this case
there is no specific absorption law like Eq. (ll), because absorption depends both on the pipe
geometry and on the flow pattern: thus, a preliminary calibration with plastic mockups that simu-
late the different flow regimes and void fractions is necessary. The cross-sectional void fraction
can also be obtained by integration of one-dimensional void fraction values measured simultane-
ously by a multi-beam device (figure 3b). Although in this case Eq. (13) holds for the single
beams, the precision of this method strongly depends on the number of beams used: since com-
mercial devices usually have 3 or 5 beams, a precise calculation of the cross-sectional void
fraction is impossible if the flow pattern is unknown.
Both the source and the detectors must be properly shielded in order to protect the operators
from exposure to radiations. On the other hand, the thickness of shielding is often considerable,
so that y-densitometers cannot be too small: for example, the most common radiation source is
137Cs, which generates y-rays at 662 keV, requiring a lead shielding 10 em thick; a less common
(and more expensive) source is 241 Arn, which releases a weaker radiation (60 keV), and needs
only a 2 mm thick lead shielding. The choice of the appropriate radiation source for a given two-
phase mixture generally represents a compromise between sensitivity (which depends on the
number of interactions in the flow) and accuracy (which depends on the number of particles inter-
acting inside the detector): if sensitivity is high, most of the interactions take place in the flow, so
that fewer particles are left for the detector, and the instrument has a low accuracy. This problem
could be solved by using X-rays instead of y-rays, because they can be produced with much
higher intensity (10 3 ..,. 104 times); on the other hand, X-ray generators are less stable and, more
important, less monochromatic, so that they are preferred for flow visualization purposes.
Radiation is generally detected by means of a scintillator coupled to a photomultiplier. The
scintillators absorb radiation and emit visible light by fluorescence; they are usually classified on
the basis of the sensitive element, which can be either solid or liquid, organic or inorganic. Or-
ganic scintillators are for instance anthracene and some polymeric materials, while inorganic
scintillators are usually ionic crystals added with impurities to increase the probability of photon
emission and reduce self-absorption, such as Nai(Tl).
Figure 4. Schematic of a photomultiplier tube.
The photomultiplier is a device that is able to convert very weak luminous pulses into electric
signals. Figure 4 shows the typical layout of a photomultiplier tube, which consists of a photo-
cathode and of a cascade of pairs of electrodes at different potentials. A small number of electrons
(smaller than the number of incident photons) is released at the photocathode, then multiplied and
focused by a series of electrodes called dynodes. The dynodes are connected to a voltage chain
produced by a high-voltage supply and a series of voltage dividers. The typical potential differ-
ence between adjacent dynodes is about 100 V, and each successive dynode is at a higher
potential than the previous one. A typical tube might have 10 dynodes: at each stage, the number
of electrons increases by a factor of the order of 5, so that the overall gain is 5 10 (about 10\
Since radiation detectors measure the radiation intensity by counting the number of particles
received in a given time interval, one cannot have "instantaneous" measurements; on the other
hand, Eq. (15) states that the relationship between the number of detected particles and the void
fraction is logarithmic: thus, the void fraction given by Eq. (15) is not a time-average value, be-
cause obviously <N> -:t. <Ln(N)>. In order to overcome this problem, one might reduce the
duration of the time interval in which particles are counted, but this reduces the instrument accu-
racy because less particles are detected; a better solution is represented by logarithmic amplifiers,
which convert the instantaneous signal into a quantity proportional to the void fraction, on which
the true time average can be calculated.
2.3 Optical Probes

Optical instrumentation is extremely common in two-phase flow diagnostics, and is used to meas-
ure several different quantities, including void fraction. In spite of their versatility, optical
instruments are essentially based on the same principle, that is, the well-known behavior of elec-
tromagnetic waves on the interface between two different media, where both reflection and
refraction occur simultaneously. The classical theory of these phenomena is described in detail in
optics textbooks, such as in Guenther (1990), and is briefly summarized here.
As it is well known, electromagnetic waves describe the propagation either in vacuum or in a
dielectric medium of the electric field and the associated magnetic field (figure 5a); for simplicity,
let's consider the electric field in a plane sinusoidal wave of frequency f = ro/2n:
E = Re lE 0ei(ax-k·x) j (16)
where k is the propagation vector.

288 V. Bertola
A 8
Figure 5. Electromagnetic waves; (a) propagation in space of the electric (E) and magnetic (B) fields; (b)
at the interface between two media of different refractive index the incident wave k1 is decomposed into a
reflected wave kR and a transmitted (refracted) wave k1 .
Maxwell's equations yield JkJ = mn I c, also called dispersion relationship, where n is the re-
fractive index of the medium where the wave propagates (in vacuum, n = 1), and c is the speed of
light. When the e.m. wave reaches the interface between two media characterized by different
refractive indexes n 1 and n2 (the plane z = 0 in figure 5b), it is decomposed into a reflected wave,
that goes back into the first medium, and a refracted (or transmitted) wave, that proceeds into the
second medium. One finds that the three waves have the same frequency, and their propagation
vectors k1. kR and kT lay in the same plane, called plane of incidence; furthermore, if 91. 9R and 9T
are the angles defmed by the direction normal to the interface and k~. kR, kT, respectively (see
figure 5b), one can obtain the well-known Snell's laws:
(}I = (}R
(17)
The reflection coefficient r is defined as the ratio between the reflected electric field and the
incident electric field; it can be calculated separately for the parallel and normal components of
the electric field with respect to the plane of incidence:
n 2 cos(}1 -n 1 cos(}T
rp =
n2 cos(}!+ nl cos(}T
(18)
n1 cos(}1 - n2 cos(}T
rN = -..!...----'--....!:...---'-
n1 cos(}1 + n 2 cos(}T
where eT can be expressed as a function of 91 thanks to the second ofEqs. (17). The quantities rp
and rN represent the fraction of the incident field reflected by the interface between the two media.
This is not the same as the fraction of reflected energy, which is proportional to the square of
these reflection coefficients (this statement is rigorous only for rp, while for rN it is an approxima-
tion).
go•
1
rp = 1
0.8
-....
I
....
0.6
0.4 a:>
0.2 n1 =1.6
0 n2 =1 rp < 0.2
-0.2
o·
o· Sc go• 1 1.2 1.4 1.6
ei n2
A B
Figure 6. Reflection coefficient; (a) parallel and normal reflection coefficient as a function of the incidence
angle; (b) contour plot of the parallel reflection coefficient as a function of the incidence angle and of the
refractive index of the second medium n2 (n 1 = 1.6).
The reflection coefficients rp and rN are plotted in figure 6a with respect to the angle of inci-
dence, for an e.m. wave passing from glass (n 1 = 1.6) to air (n2 = 1). When 91 is larger than a
critical value Elc, both rp and rN become complex numbers with a modulus equal to 1: thus, the
amplitude of the reflected field is the same as that of the incident field, and the only difference is
in the phase. For 81 < Sc both the coefficients drop to about 20%: in terms of reflected energy, this
corresponds to 4% of the incident energy. Figure 6b shows the contour plot of the parallel reflec-
tion coefficient rp as a function of 91 and of the refractive index of the second medium n2,
assuming that the first medium is glass with n = 1.6; the parallel reflection coefficient rN is analo-
gous.
The optical probe is an instrument that allows to distinguish two fluids of different refractive
index, thanks to the abrupt change of the reflection coefficient that can be achieved when n2
changes with constant Elh as shown in figure 6b for 91 greater than about 30°. The probe consists
of a glass optical fiber (refractive index n 1 ~ 1.6) with a diameter of about 100 11m, connected at
one end to an infra-red source and to a photo-diode, whereas the free end is placed inside the
fluid; this technology has been known for more than 30 years (Hinata, 1972; Jones and Delhaye,
1976; Cartellier, 1990; Cartellier and Achard, 1991 ). Figure 7a shows the working principle of a
typical single-fiber optical probe: when the light emitted inside the optical fiber reaches the tip, it
is completely reflected back if the free end is immersed in air, while the reflection coefficient
drops to almost zero when the probe is immersed in water; the reflected light is converted by the
photo-diode into an electric signal, which ideally consists of two voltage levels, and corresponds
to the phase density function. The difference in the signal level for the liquid and the gas phases
has been found to vary with the void fraction and phase velocities, resulting in threshold or dis-
crimination level uncertainties, as shown for example by Abuaf et al. (1978).
290 V. Bertola
light source
detector
Ni-Cr
in gas total in liquid no (mirror)
A reflection reflection B
Figure 7. Optical probes; (a) working principle; (b) conical probe; (c) probe; (d) U-shaped probe, no longer
in use.
Inside optical fibers the inclinations of light beams (which correspond to the propagation vec-
tors) are statistically distributed around their most probable direction, coincident with the fiber
axis; for simplicity, we can assume that all the beams have this direction: thus, we can also define
a single angle of incidence 8h which is determined only by the geometry of the free end. For
instance, for a conical tip 81 is equal to the cone half angle. The choice 81 = 45° represents an
optimal solution to distinguish gases from liquids with a glass optical fiber (n 1 :::::: 1.6), because in
this case we have that total internal reflection occurs for n2 < 1.15 (this value can be determined
graphically with the aid of figure 6b, or calculated imposing Or = 90° in the second of Eqs. 17)
while for n2 > 1.15 reflection is not significant. Thus, this system can properly distinguish gases at
low pressure (n2 :::::: 1) from most liquids, which generally have n2 > 1.15 (for water, n2 :::::: 1.33; for
kerosene, n2 :::::: 1.44; freons have n2 :::::: 1.25).
In order to be detected by the photo-diode, the beam must be deflected of 180°; since OR= 8],
this can be accomplished by means of two total reflections with 8 1 = 45°: figure 7 shows three
possible geometries of the probe free end. The conical tip (figure 7b) is the simplest geometry,
which can be obtained for instance by etching a cleaved fiber with a corrosive liquid; due to the
very small dimensions (the cone tip may be as small as 20 fJ.m), manufacturing conical tips is
extremely difficult, and finding two probes which have exactly the same shape is almost impossi-
ble: Cartellier and Barrau (1998a) showed that probes supplied by different manufacturers have
different responses in the same operating conditions, so that a careful calibration of each probe is
necessary. The cleaved probe (figure 7c) has a flat surface inclined at 45° with respect to the fiber
axis; the end of the fiber is covered with a reflecting coating (e.g. a Ni-Cr paint), so that after a
first total internal reflection the light beams hit the coating and are reflected back to the active
surface, where they undergo a second total internal reflection back to the photo-diode. This probe,
described in detail by Fordham et al. (1999a,b), can be manufactured very easily by cutting at 45°
a bundle containing several hundreds of fibers. Another possible geometry is represented by the
U-shaped probe (figure 7d), in which the double total internal reflection is obtained by bending
the fiber of 180°; although this configuration has been very popular in the past years, its perform-
ances cannot be compared with those of the conical probe, and nowadays it is no longer in use.
~
<OJ
-u S1
I
I
'
:'
bubb le
surface
:'
:I I
0.
'$ I
0 I :'I
S2 I
I
I
' I
I
liqu id
fi lm
~
Figure 8. Dynamic response of an optical probe to the transition from one phase to another; the detail
shows the liquid film still wetting the probe surface after the tip has entered a gas bubble.
The dynamic response of the optical probe to a change of phase is schematically depicted in
figure 8, which describes the passage of a gas bubble through a probe initially surrounded by
liquid. In the liquid phase the voltage output has a constant value corresponding to no reflection;
when the probe enters the gas phase, the reflected beam activates the photo-diode, increasing the
voltage output to the higher value. The growth of the output value is not instantaneous, but re-
quires a time ~t~. because the smaller contact angle of liquids on glass causes the probe tip to be
covered by a thin liquid film, which is dried out only when the tip has already advanced into the
gas phase; as the film becomes thinner, the reflection coefficient grows, until the output reaches
the maximum value. Since wetting phenomena are extremely complex (see e.g. the review article
by Bonn and Ross, 2001), the time delay is sensitive to a great number of parameters (probe ge-
ometry, orientation with respect to the bubble direction, impurities in the fluid, manufacturing
procedure, etc.), and can be determined only empirically. A similar phenomenon occurs when the
probe passes from the gas to the liquid: the output value drops only after the tip has entered the
liquid phase. In this case, the time delay ~t2 is shorter than ~t 1 , because the liquid tends to spread
over the probe surface, accelerating its wetting. These signal transients make the proper identifi-
cation of the bubble 'front' and 'back' interfaces difficult. If the bubble velocity is not too small,
it is inversely proportional to the time delay ~t 1 (Cartellier and Barrau, l998a): thus, the quantity
L = u ~t 1 (the so-called latency length) is a constant, and can be used to characterize the probes
quantitatively.
In order to discriminate the two levels of the output signal S, a threshold value S0 could be
chosen such that one phase corresponds to S < S0 and the other to S > S0 . Due to the time delays
~t 1 and ~t 2 this method is not very precise: as one can verify qualitatively on the plot in figure 8, a
value of S0 close to the lower value of the signal (liquid phase) allows to detect with precision the
transitions from liquid to gas, whereas the detection of passages from gas to liquid is affected by
an error; similarly, if S0 is close to the higher value of the signal (gas phase) only gas/liquid inter-
faces are detected precisely; if S0 has an intermediate value between the two levels, both
liquid/gas and gas/liquid interfaces are missed.
292 V. Bertola
I\
'
r
'' r
I
I
' I
I
'
I
\
\ r
,(
'
''
\
' \
''
I
\ I
I
'\
\
Figure 9. Cleaved optical probe: total internal reflection is never possible.
A better discrimination technique uses two threshold levels S 1 and S2 (figure 8), the former for
gas/liquid and the latter for liquid/gas interfaces: the passage from liquid to gas corresponds to S
= S2 , while S = S 1 denotes the passage from gas to liquid. Past experience shows that the probe
orientation with the flow direction has little effect on the measurement accuracy, provided that the
probe support does not affect the flow past the tip (e.g. Sene, 1984).
Unlike the probes described above, the cleaved probe (figure 9) is not based on the phenome-
non of total internal reflection. In fact, this phenomenon occurs when the reflection coefficient
equals 1, which is possible only if 8 1 is not too small, as one can verifY for rp on the plot of figure
6b (for a glass probe 8 1 should be at least 30-35°). In the cleaved probe, beams are almost perpen-
dicular to the probe end (8, ~ 0°), so that the reflection coefficient is always smaller than 0.2,
corresponding to a fraction of reflected energy smaller than 4%, and total internal reflection never
takes place; this kind of reflection, which is known as Fresnel reflection, is the cause of the faint
images reflected by window panes, for instance. Eqs. (17) and (18) show that the reflection coef-
ficients change in a continuous fashion with the refractive index n2 of the second medium: thus,
the cleaved probe can be used to distinguish fluids with different refractive indexes, even if they
cannot de discriminated by conical probes. This happens, for instance, in liquid-liquid flows, in
which all the components have n 2 > 1.15, so that conical probes give a constant output signal for
both phases.
Hamad et al. (1997) built a cleaved probe from a glass fiber with n 1 = 1.48, which could be
used to distinguish the different phases of a water/kerosene flow. The optical probe cable had a
100 11m core diameter and a 140 11m cladding diameter; the cladding provided a constant step in
the refractive index at the core-cladding interface, which guided the light inside the fibre. The
current through the photo-diode was directly proportional to the strength of the light received.
The energy fraction reflected by this probe was 0.28% in water (n2 = 1.33) and 0.024% in kero-
sene (n2 = 1.44), and grew to 3.5% in air. As one can see from these values, only a very small
proportion of the transmitted light is reflected and delivered to the photo-diode receiver, resulting
in a very small operational signal for phase detection: consequently, considerable attention must
be paid to the signaVnoise aspect of the system.
If the diameter of the probe is much smaller than the size of the dispersed phase, the use of a
flat-tipped probe has distinct interface detection advantages compared with the use of a conical
probe: in fact, surface tension effects will cause different responses for the two probes. When first
contact occurs between an interface and a flat tip, with the contact angle being very close to zero
(Farrar and Bruun, 1989), the interface will virtually instantaneously deform so that the whole of
the tip comes into contact with the interface, causing a steep change in the probe signal. On the
contrary, when using a conical probe a meniscus will be formed, which will gradually creep up
the conical tip as the bubble or drop interface moves past the tip, which results in a much longer
transition time (Bruun 1995).
Another technique to distinguish the phases in liquid-liquid mixtures consists in adding a fluo-
rescent dye to one of the two fluids, and excite fluorescence with a laser light source at the
appropriate wavelength; some liquids, such as light crude oil, have a little natural fluorescence,
and need no additives. Ramos et al. (200 1) applied this technique to distinguish the components
of a three phase mixture (oiVwater/gas) by a combination of reflectance and fluorescence methods
(Wu et al., 2000). The signal reflected due to the contrast in refractive index provided by gas is
directly detected by a photo-diode, while the fluorescence signal is separated from the excitation
signal by a long-wavelength-pass filter and detected by a second photo-diode to provide the de-
tection of oil.
2.4 Electrical Impedance Methods

Electrical impedance is the measure of the degree to which an electric circuit resists a flow of
electric current when a voltage is impressed across its terminals: in particular, impedance is the
ratio of the voltage impressed across a pair of terminals to the current flow between those termi-
nals; the SI unit for this quantity is the Ohm (Q). In direct-current (DC) circuits, impedance
corresponds to resistance; in transient or alternating current (AC) circuits, impedance is a function
of resistance, inductance, and capacitance. Inductors and capacitors build up voltages that oppose
the flow of current: this opposition, called reactance, is combined with resistance to find the im-
pedance. The reactance produced by inductance is proportional to the frequency of the alternating
current, whereas the reactance produced by capacitance is inversely proportional to the frequency.
In AC circuits, both the voltage V and the current I can be expressed as rotating vectors at a fre-
quency ro/2n and phase <p; with this representation, electrical impedance is a complex quantity,
where the real part is the resistance R, and the imaginary part the reactance X:
V Vei(OJI+9"J) V ·Li V V
Z=-= ( ) =-e' rp =-cosAqHi-sinAtp=R+iX (19)
I Ie' OJI+rp2 I I I
Electrical impedance is also a measurement of how electricity travels through a given mate-
rial: every material has different electrical impedance determined by its molecular composition. In
case of a heterogeneous material, impedance depends both on the composition and on the spatial
distribution of the different components. Thus, the electrical impedance exhibited by an assigned
volume of a two-phase mixture (in particular, gas-liquid) depends both on the void fraction and
294 V. Bertola
on the flow pattern: this provides a way to measure the void fraction once the response for differ-
ent flow patterns is known, for instance by means of a preliminary calibration.
As shown by Ceccio and George (1996) in their extensive review analysis on the impedance
method, there are many different possibilities to arrange a system of electrodes for void fraction
measurement purposes. Regardless of the electrode arrangement, the main problem to tackle
when employing the impedance technique is to obtain the proper calibration curve since the probe
response to void fraction variations depends upon phase distribution within the volume where the
measurement is taken. The problem is described by Laplace's equation of the electrical potential
with the proper boundary conditions:
(20)
Depending upon the geometry of the probe and the form of the interface between the phases,
Laplace's equation can be solved either analytically or numerically. The probe response is gener-
ally nonlinear, and it depends on the electrode geometry (dimensions, shape, distance). As a
consequence, one should know a priori the flow pattern established inside the system under ob-
servation in order to apply the proper calibration curve to the measurements.
According to the physical properties of the mixture components and to the frequency of the
current applied to the electrodes, the electrical impedance may exhibit a resistive or a capacitive
behavior: in particular, for air-water mixtures the resistive component dominates in the frequency
range between 10kHz and 100kHz, while for higher frequencies (typically 1 MHz) the capaci-
tive component is prevailing.
Electrical resistance. These instruments are based on the measurement of the electrical resistance
R of a two-phase mixture, where at least one of the fluid is conductive. Since the resistance of
conductive fluids is very small, the electrical conductance G = 1/R is often used instead. In gen-
eral, electrical conductance measurement devices require simple instrumentation and provide
reliable results; the main limitation of this kind of probes is that the electrodes must be in contact
with a conductive fluid, so that the probe geometry must be designed taking into account the flow
pattern; furthermore, the measured conductance is strongly influenced by undesired factors due to
the electrode fouling or to temperature changes. The measurement of the unknown resistance is
usually done by means of a Wheatstone bridge configuration (figure 10), which provides an easy
way to compensate bias errors due for instance to a thermal drift of the water conductance. As
shown in figure 10, if the bridge is initially balanced (R1R3 = R2~) the changes of the resistances
placed on the same side of the bridge with respect to the power supply have different signs in the
expression of the output voltage Vout. so that if they both have the same bias error this will not
affect the measurement. Practically, these two resistances are represented by the measuring probe
and by a slave probe, which is immersed in water kept at the same temperature as the water in the
flowing mixture.
In principle, conductance measurements can be performed indifferently by means of DC or
AC current; since DC current applied to a conducting fluid induces electrolytic phenomena, AC
current in the range 10 +100kHz (where the reactive component of impedance is not significant)
is used to avoid the corrosion of the electrodes. A remarkable exception is represented by local
probes such as those described by Kim et al. (2000): they used gold-plated electrodes to increase
the life of the sensor.
Figure 10. Wheatstone bridge configuration for resistance measurements with error compensation.
There are many different possibilities to arrange a system of electrodes for either local or inte-
gral void fraction measurement, as described in a comprehensive review of the technique by
Hewitt (1978). A system of electrodes assembled to constitute an impedance probe has to meet
several functional and practical requirements. First of all, the probe must ensure an effective con-
finement of the electrical field in order to minimize the edge effects and enhance the spatial
resolution. Furthermore, the shape of the sensor should produce a linear response to the variations
of the observed quantity. Finally the electrodes have to be easily manufactured and installed in
p1pes.
Local electrical conductance sensors were first proposed by Neal and Bankoff (1963). These
sensors (figure lla) are one of the most widely used measurement techniques in obtaining local
two-phase flow parameters, such as the phase density function and the interfacial area concentra-
tion. When immersed in a mixture of one conducting and one non-conducting fluid, the
characteristic rise/fall of conductance signals can be acquired as the two phases pass through the
sensor. Depending on the application, single-sensor probes, double-sensor probes (Kataoka et al.,
1986) or four-sensor probes (Kim et al., 2000) can be used. Single-sensor probes consist of two
coaxial metallic electrodes, divided by a layer of insulating material; the diameter is generally
smaller than 1 mm.
The studies by Brown et al. (1978), Karapantios et al. (1989), Koskie et al. (1989), Ruder and
Hanratty ( 1990) concern the use and calibration of parallel wire probes. As shown in figure 11 b,
these sensors consist of two parallel wires perpendicular to the flow direction; wires are generally
in copper or platinum, with a diameter of about 0.2 mm. The electrical conductance between the
two wires is proportional to the wetted length: therefore, these sensors are useful to determine the
liquid layer thickness in separated flows (stratified, annular), from which the cross-sectional void
fraction can be inferred. Wire probes are also used to detect the passage of liquid slugs in inter-
mittent flows, because when a slug passes through the sensor the output signal switches to a value
different from that corresponding to the stratified region: in particular, slug frequency can be
estimated by counting the number of slugs per unit time or by Fourier analysis of the output sig-
nal, while slug velocity can be inferred by cross-correlating the signals of two probes placed at a
known distance along the pipe.
296 V. Bertola
I J
flow
external electrode
lnsulant
l nn~r e>k!ctrod~
A B c
Figure 11. Electrical conductance probes; (a) local measurements (phase density function) can be per-
formed with two coaxial electrodes divided by insulating material; (b) in wire probes, electrodes are
parallel metallic wires; (c) in ring probes electrodes are flush mounted on the pipe wall.
Probes with flush-mounted electrodes provide a reliable non-intrusive method for integral
(cross-sectional or volume averaged) void fraction measurements. Flush-mounted ring electrodes
were employed by Asali et al. (1985), Andreussi et al. (1988), Tsochatzidis et al. (1992), Fossa
( 1998); a typical arrangement is represented by two metallic rings annealed in the pipe inner wall,
as shown in figure llc. The bases for integral measurement devices were given by Coney (1973),
who described the theoretical behavior of flat electrodes wetted by a liquid layer. In particular, he
obtained that the conductance between plane electrodes of length L covered by a liquid layer of
thickness h is given by:
G = GLL K(m) (21)

K(l-m)
where:
· h2(1r
Sill -
2h
s) (22)
m-
- sinh2[ 7r(s ;hDe)]
---;::-+-~=
In Eqs. (21) and (22) GL is the conductance of the liquid, sis the electrode width and De the dis-
tance between the electrodes. Therefore, from an impedance measurement one can calculate the
liquid layer thickness.
EEl-$-BBB
A B
Figure 12. Optimized geometries of flush-mounted electrical conductance probes; (a) half-ring probes with
variable width (black) between guard electrodes (gray); (b) ring probes with variable axial distance (cour-
tesy of Marco Fossa).
Andreussi et al. (1988) showed that Coney's solution holds even for a cylindrical geometry,
provided a proper defmition of the geometrical parameters is used: in particular, the liquid layer
thickness must be replaced by the ratio between the cross-sectional area of the liquid and the
wetted perimeter. Ring electrodes have also been considered by Tsochatzidis et al. (1992), who
theoretically solved Laplace's equation in cylindrical coordinates for the ring-probe response to a
conducting annulus of thickness h.
When ring electrodes are employed to determine the void fraction in uniformly dispersed two-
phase systems (e.g. bubble flow or fluidized beds), the apparent conductance of the two-phase
mixture is expected to be proportional to the mean volumetric liquid fraction (1-e). Bruggeman
(1935) suggested the following expression:
(23)
Eq. (23) well agrees (the maximum difference is of a few percent fore < 0.3) with the solution
obtained by Maxwell (1882):
G = GL 2(1 - c) (24)
2+&
For other electrode configurations (e.g. plates), no theoretical solutions have been developed:
a few data are available in the form of experimental probe responses obtained with reference to
the uniformly dispersed flow pattern. This is the case of the probe employed by Ma et al. (1991)
and later by Costigan and Whalley (1997) which consists of a pair of measuring half-rings located
along the pipe circumference; the electrodes are placed between two pairs of guard electrodes
(half-rings, also) at the same potential as the corresponding measuring electrodes.
298 V. Bertola
Among the various configurations proposed in the literature, the flush-mounted ring and half-
ring probes seem to have the best performances. The main difference between these families is
that the electrical field is oriented mainly along the pipe axis in ring probes, while in half-ring
probes it is oriented perpendicular to the axis. Starting from these reference configurations, the
electrode geometry can be modified in order to achieve better performances in terms of response
linearity and spatial resolution. Fossa and Devia (1999) evaluated the response of the probes
under different flow conditions by solving Laplace's equation with a finite-element commercial
code (Ansys 5.5). On the basis of numerical simulations, two new geometries have been devel-
oped, which are characterized by variable electrode width (figure 12a) or variable distance
between the electrodes (figure lb). Preliminary tests (Devia and Fossa, 2003) confirm that these
geometries greatly improve the probe response, both in terms of linearity and in terms of dynamic
response to step changes in the void fraction.
Electrical capacitance. Electrical capacitance sensors provide a fully non-intrusive way to meas-
ure the average void fraction over a given volume of a two-phase mixture: in fact, unlike
resistance sensors, the electrodes need not to be in contact with the fluids. Another advantage is
that, unlike conductivity, capacitance is not much sensitive to temperature drifts and does not
depend on the liquid ionization. In general, the measured capacitance represents the amount and
the configuration of the phases within the sensor; for a gas-liquid mixture the capacitance C can
be expressed as:
(25)
where ~ and ~L are the dielectric constants of the two fluids, E the volume average void fraction
and K is a factor which takes into account the sensor volume and geometry, as well as the flow
pattern. Capacitance measurement is possible by integrating over the sensor volume the well-
known relationship C(x) = dq/dE, where q is the electric charge and the potential Eisa solution
of Laplace's equation (Eq. 20). For symmetrical electrodes surrounding a homogeneous dielectric
the analytical solution of this problem exists, and can be easily extended to a multi-layer cylindri-
cal geometry (which is the case of annular flow in vertical pipes). If the distribution of the fluids
inside the sensor volume is not axisymmetric, no analytical solution exists, so that only approxi-
mate or empirical solutions are possible.
Capacitance sensors have been widely investigated, for instance by Merilo et al. (1977), Ma-
suda et al. (1980), and Heerens (1986) among others. Unfortunately, especially for gas-liquid
flow measurements capacitances are very small, generally in the range of 0.1 to 10 pF: thus,
proper shielding against stray capacitances and a good signal to noise ratio are mandatory. Stray
capacitance can occur between circuit wires, wires and the chassis, electronic components, and
even between the sensor terminals if they are not properly designed.
A variety of electrode configurations has been designed by a number of researchers in the past
few decades, ranging from flat plate, concave, helical and multiple helical wound (figure 13),
where either the electrodes were in contact or isolated from the fluid (Huang et al., 1988). Among
all of these configurations, best performances were obtained with the helical wound sensor (figure
13c), described for instance by Gregory and Mattar (1973), Albouelwafa and Kendall (1979) and
Geraets and Borst (1988), and with the concave plate sensor (figure 13d), described among others
by Elkow and Rezkallah (1996, 1997) and Lowe and Rezkallah (1999).
•
_() I
•
A B
c D
Figure 13. Configuration of electrodes for electrical capacitance probes; (a) flat plate electrodes; (b) multi-
electrode probe; (c) helical wound probe; (d) concave plates probe.
For the helical wound sensor, empirical design criteria are:

w
- = 0.1 + 0.136 (26)
p
and
.!!... = 1.83 + 3.66 (27)

D
where w is the width of the electrode, p is the pitch and D is the outside diameter of the pipe.
Besides an undoubted complexity, which requires an extremely careful design and construction,
the helical wound sensor has two intrinsic limitations: the presence of stray capacitance between
the sensor electrodes, which cannot be completely eliminated by guard electrodes, and nonlinear
response at the ends of the sensor, which is due to the different length of the electrodes and makes
the instrument more sensitive to the flow pattern.
The concave parallel plate sensor represents a great improvement over the helically wounded
sensor, provided guard electrodes are placed at the sensor ends, and appropriate shielding against
stray capacitances with other electronic equipments is provided. The sensor length is a key design
parameter: in fact, the measured capacitance is proportional to the sensor volume, so that stray
capacitances are less important for long sensors; on the other hand, if the sensor is too long, the
void fraction fluctuations cannot be detected. Typical values for the sensor length range between
2 and l 0 pipe inner diameters.
300 V. Bettola
voltage
guard
screen
active
CA-G
ground
A B
Figure 14. Stray capacitance; (a) schematic description of stray capacitances between the electrodes and
the shielding; (b) active-guard circuit for stray capacitance compensation.
Possible electronic schemes for this kind of instrument are discussed in detail by Huang et al
(1988). While an acceptable resolution (0.01 pF) can be easily achieved with standard electronic
components, the major problem is represented by the stray capacitance between the electrodes
and the screen (figure 14a), which may be of the same order of the capacitance to be measured.
Moreover, if one of the electrodes is grounded (which is a common practice when accurate, high-
frequency measurements are required), the measurement error is not distributed homogeneously
inside the sensor's volume. This problem can be solved thanks to the so-called "active guard
method", described schematically in figure 14b: in this case, the stray capacitance between the
live electrode and the screen is replaced by the electrode-guard and the guard-screen capacitan-
ces; thus, if the guard is driven to the same potential as the live electrode (which can be easily
done by a commercial voltage-follower), the electrode-guard capacitance behaves like an open
circuit, so that the measurement is no longer affected by stray capacitances.
3 Flow Visualization
Two-phase flow visualization techniques are extremely important in two-phase flow modeling,
because they allow one to quantify the distribution of the phases inside a controlled volume or
over a given cross-section: thus, they provide essential information about the flow pattern. Such
techniques are quite different from those used to visualize single-phase flows, which are con-
ceived to be used in a homogeneous medium. In particular, the complex interface between the
two phases causes multiple scattering of visible and near-visible radiation, on which several visu-
alization techniques, such as laser interferometry, are based. Therefore, like for void fraction
measurements, the instruments must be able to discriminate between some physical property
which is different for the two phases (e.g. attenuation coefficient, electrical impedance, etc.). A
common feature to these devices is that, in addition to the instantaneous or time-average distribu-
tion of the two phases in a given geometrical domain, they also give the value of the space-
averaged void fraction inside the same domain, which comes from simple integration of the phase
density function values.
3.1 Radiography
Radiography provides a very efficient tool for investigations in the field of non - destructive test-
ing as well as for many applications in fundamental research. A beam penetrating a specimen is
attenuated by the sample material and detected by a two dimensional imaging device. The image
contains information about material and structure inside the sample because the beam is attenu-
ated according to the basic exponential law of radiation attenuation.
X-ray radiography has a long tradition as an important tool for medical applications as well as
in the field of non destructive testing. While penetrating matter, X-rays interact with the electron
clouds of the atoms in the object. Therefore the attenuation ofX-rays depends on the charge den-
sity of the electron cloud and increases with the atomic number of the sample material. In general,
as the atomic number increases, X-ray attenuation increases for lower X-ray energies and/or
higher density materials.
Neutron radiography is similar in principle to X-ray radiography, and is complementary in the
nature of information supplied. The interactions of X-rays and neutrons with matter are funda-
mentally different, forming the basis of many unique applications using neutrons. While X-rays
interact with the electron cloud surrounding the nucleus of an atom, neutrons interact with the
nucleus itself. Thus, the total neutron cross-section subject to the neutron energy depends on the
properties of the nuclei and varies from element to element, even from isotope to isotope. Con-
trary to X-rays, neutrons are attenuated by some light materials, as i.e. hydrogen, boron and
lithium but penetrate many heavy materials (e.g. lead, iron, chromium, etc.). Neutrons are able to
distinguish between different isotopes, and neutron radiography is an important tool for studies of
radioactive materials.
Neutron radiography is a powerful technique for the investigation of two-phase flows with a
growing number of applications (Hibiki et al., 1995; Mishima and Hibiki, 1996). In particular, it
is useful when the operating conditions do not allow to place the instrumentation in contact with
the fluids and the insertion of a transparent section is impossible; this happens, for instance, in
thick walled pipes containing fluids at high pressures and temperatures. An example of applica-
tion of neutron radiography to boiling flows can be found in Kureta et al. (1999).
3.2 Tomography
The word tomography denotes a number of experimental techniques that allow one to visualize a
two-dimensional image of a given cross-section ofthe measured object (figure 15). Tomographic
techniques are based on the simultaneous measurement of a physical property in several positions
on the measurement plane. Such techniques, which are commonly used in the medical and mate-
rial science fields, have been recently applied to the process industry.
Process tomography has been developing rapidly in recent years, and now has come to the
stage of industrial applications, as it is reported in several exhaustive review papers (Beck and
Williams, 1996; Chaucki et al., 1997; George et al., 1998; Mewes and Schmitz, 1999). Tomo-
graphic techniques provide novel means of visualizing the internal behavior of two phase flow in
industrial processes, such as gas/liquid three-component flows in oil pipelines (gas/oiUwater),
gas/solid flows in pneumatic conveyors and processes of mixing or separation in plant vessels.
These systems can provide more valuable information for measurement, on-line monitoring and
control of industrial processes than can traditional equipment, as shown among others by Scott
and Williams (1995), Xie et al. (1995) and by Thorn et al. (1999).
302 V. Bertola
f{x,y) f(x,y)
integral
measurements
)"
y Ireconstruction
algorithm
Figure 15. Principle of process tomography and schematic of the operating sequence.
The tomographic measurement is performed in two steps, which are illustrated schematically
in figure 15. In the first step, a number of integral, independent measurements ~n(~n) of the prop-
erty field f(x,y) are taken in the measurement plane by the tomographic sensor. After
measurements have been completed, the local properties of the flow are calculated by means of
the so-called reconstruction algorithm, which returns a distribution function f(x,y) of the meas-
ured quantity; of course f(x,y) might not be identical to f(x,y) . Figure 16a depicts a typical
arrangement for integral measurements by X-ray or y-ray tomography: the measurement object,
placed on the plane (x,y), is projected in each direction thanks to several beams, that are attenu-
ated according to its composition. Linearly independent projections %(~n) can be obtained by
rotating the measurement device or by using more beam sources. The integral measurements
implicitly contain the information about the measured property f(x,y): the larger the number of
measurements, the more detailed will be the reconstruction. In practice, the number of independ-
ent projections cannot be smaller than three in order to avoid ambiguities: in the example of
figure 16b, the two orthogonal projections may correspond to any of the two couples of cylinders
(1) and (2), placed on the diagonals of the measurement plane.
Measurement devices are based on the interaction of the multiphase flow with electromagnetic
waves, particle radiations or acoustic fields, as shown in detail by Mewes and Schmitz (1999);
according to the type of interaction between the sensor and the flow, they can be classified into
three categories:
- transmission techniques: the flows changes some physical property of the emitted field,
which can be measured by the sensor receiver; such properties include e.g. the polariza-
tion of e.m. waves or the intensity of electric and acoustic fields (see e.g. Yang, 1996);
1 2
2 1
A B
Figure 16. Integral projections in process tomography; (a) example of linearly independent integral meas-
urements in three different directions; (b) schematic description of ambiguities arising when n < 3.
- emission techniques: the sensor measures the radiation emitted by the material inside the
measurement cross-section, which can be either spontaneous or induced by the sensor it-
self; these techniques include infrared termography, magnetic resonance imaging
(Gladden, 1995; Le Gallet al., 2001) and positron emission tomography (Parker et al.,
1994);
- deviation techniques: the flow changes the direction of the radiation emitted by the sen-
sor; these techniques include reflection, refraction (deviations are due to an interface),
diffraction (deviations occur in a continuous medium), scattering.
Reconstruction algorithms are used to extract the information about the measurement object
from the integral measurements. Their theoretical bases have been developed in the last decades,
(Herman, 1980; Natterer, 1986; Kak and Slaney, 1988). The algorithms are usually divided into
inversion techniques, in which measurements are mathematically inverted after some pre-
processing of data, and series expansion techniques (Censor, 1983), for which the discretization
of the measured function f(x,y) is required. In series expansion techniques, the sensor is described
by a linear system of equations expressing the (discretized) measurements as a function of the
(unknown) discrete values of f(x,y), which can be reconstructed by solving the system. If the
number of measurements equals the number of unknown elements of the measured property, the
direct analytical inversion of the measurement field is possible; examples are direct inversion
techniques such as convolution, 2-D and 3-D Fourier transforms, etc. (Mersereau, 1976). On the
other hand, in process tomography the number of projections is generally small: thus, iterative
techniques are used in order to improve the quality of the final reconstruction.
304 V. Bertola
f'(x,y)·..
+.... "' A<Dn ... optimizer ....
n -
(I)'
n math. descr. ....
Figure 17. Feedback loop of iterative reconstruction methods.
The output property distribution f(x,y) obtained from the integral measurements <l>n(~n) is in-
troduced into a mathematical model of the sensor, that yields a new set of measured values
%'(~0), which are supplied back to the reconstruction algorithm until convergence is reached, as
shown schematically in figure 17. Examples of these methods are given for instance by Rowland
(1979) and Herman (1980).
Tomographic measurement systems are required to have a high temporal resolution, in order
to detect the evolution of transient flows; moreover, a sufficiently high spatial resolution is neces-
sary in order to recognize the flow structure. Of course, these characteristics are strongly
influenced by the working principle of the sensor: in particular, classical techniques such as X-ray
or nuclear magnetic resonance tomography have a high spatial resolution, suitable for biomedical
applications, but a rather low time resolution (1 + 500 s), so that they cannot be used in transient
multiphase flows of process industry. For these applications, better solutions are electrical capaci-
tance or acoustic tomography, which allow a high temporal resolution together with a low (but
still acceptable) spatial resolution.
Electrical conductance tomography. This technique has been recently developed (Xie et al.,
1995; Reinecke et al., 1998), and probably represents the most efficient, low-cost method for two-
phase flow diagnostic, where one of the fluids is electrically conductive. As it is schematically
depicted in figure 18a, the sensor consists of three sets of parallel wires placed in three adjacent
parallel planes perpendicular to the flow direction, so that they create a mesh through which the
two-phase mixture flows. The distance between two wires in the same plane is of the order of 5
mm, while the axial distance between two planes is about 1.5 times larger; stainless steel wires
with a diameter of 0.1 mm were used, so that the free cross-sectional area through the sensor is
greater than 95%, and consequently the additional pressure drop is not significant. This sensor,
though, is intrusive, and might somewhat alter the flow structure; furthermore, in the three-array
configuration measurements cannot be performed over the whole circular cross-section, but only
in the exagonal area depicted in figure 18a.
The working principle of the sensor is based on the conductance measured between two paral-
lel wires, which is a function of the electrical properties and of the distribution of the two phases.
Depending on the flow pattern, the liquid bridges the gap between each couple of wires in several
positions, as shown in figure 18b: the liquid bridges, which have different lengths Li, can be de-
scribed by a set of parallel ohmic resistors.
Arra 2
MULTIPLEXER
Pipe cross-section
A B
Figure 18. Electrical conductance tomography; (a) arrangement of the electrical conductors in the meas-
urement cross-section: note that the axial distance between two arrays is of the same order of the distance
between two wires; (b) schematic representation of the conductive fluid bridges of length L; between each
couple of wires, which are equivalent to a succession of parallel electrical resistances (conductances).
Neglecting the conductance of the gas, the measured conductance is:
L:-1 =-1-LL;
N N
_!_= (28)
R i=l R; LRw i=l
where Rw is the electrical resistance of the liquid, and L is the total length of the wire in contact
with the mixture. The ratio of the wetted length to the total length equals the one-dimensional
liquid holdup (l-s 1). The electrical conductance represents the integral measurement to be sup-
plied to the reconstruction algorithm; each set of wires provides a linearly independent projection
of the measurement plane. If conductance measurements are performed simultaneously, the con-
ductance between two wires will be affected by the electric field created by the other active wires:
thus, a multiplexing device that scans in rapid succession all the couples is necessary. Of course,
such sequential measurements require a time shift correction strategy in order to be considered as
if they were simultaneous.
The parallel wire electrical conductance sensor allows a time resolution of about 0.01 s, which
means that cross-sectional images of the phase distribution can be acquired at a rate of about I 00
Hz. Three dimensional representations of the flow structure are also possible: in fact, by stacking
the reconstructed images, an Eulerian representation of the phase distribution can be obtained.
306 V. Bertola
TRANSMITTER
transmitter
..[],.flow
,_ waveguides
~"' """"
..........
\ • •
0:::
w
I
:::0
m
'-\
~
~ 0
w m
0
w
\ <
m
controlled
volume
., ~~
~
0::: r...__
:::0 receiver
waveguides
"""""
A B
Figure 19. Ultrasound acoustic tomography; (a) configuration of the acoustic waveguides in the two-phase
flow; (b) detail of the arrangement of the transmitter and the receiver waveguides in the controlled volume
of fluid.
Acoustic tomography. These systems, based on the measurement of the acoustic conductivity of
gas-liquid mixtures, have been recently proposed as a reliable diagnostic technique for two-phase
flows at high pressures and temperatures; examples are reported by Melnikov and Nigmatulin
(1994), Schlaberg et al. (1996), Melnikov and Kontelev (1999). The ultrasound acoustic waves
generated by piezoelectric transducers are irradiated in the mixture by means of an acoustic
waveguide (for instance, metallic rods or thin-walled tubes), and reach another waveguide con-
nected to a receiver, which converts the acoustic fluctuations into an electrical signal. If the fluid
between the transmitter and the receiver waveguides is a liquid, the acoustic waves attenuation is
not significant, while if the fluid is a gas the ultrasound fades and does not reach the receiver; in
fact, the acoustic energy transmitted by the fluid is proportional to the acoustic impedance Z = pc,
where pis the fluid density and c the sound velocity: for water, Z:::: 1.4 x 106 kg m· 2s· 1 whereas
for air at atmospheric pressure we have that Z:::: 4 x 102 kg m·2s· 1.
Figure 19 shows the scheme of the acoustic sensor proposed by Melnikov and Kontelev
(1999): the irradiation of ultrasound waves is carried out by means of separate, parallel
waveguides, along which the free ends of the receiver waveguides are located at a distance of 1
mrn. The transmitter and the receiver waveguides create an orthogonal grid that covers the whole
cross-section of a pipe, and allow a spatial resolution of about 1 em.
Acoustic tomography has two important advantages with respect to other tomographic tech-
niques: it can be used in extreme operating conditions (temperature up to 350°C and pressure up
to 20 MPa), and there are no specific requirements about the fluid properties, whereas for electric
conductance tomography at least one of the fluids must be conductive. The high resistance of this
kind of sensors is due to the fact that only the waveguides are placed in contact with the fluids,
while all the electronic components and conductors are kept outside, in a safer environment. The
main drawbacks are the relatively low temporal and spatial resolution: the former is related to the
path of acoustic pulses and to signal processing, while the latter depends on the size of the
waveguides, which is of the order of 1 mm, and on the distance between waveguides, which must
be optimized in order to avoid interferences.
3.3 Time-Average Flow Visualization

While the instantaneous visualization of the whole flow field gives the most detailed information
about the flow structure, deducing from it the overall behavior of the flow is often difficult: there-
fore, sometimes a time average description of the flow structure can be much more useful. One
such representation is provided for by the local void fraction distribution, which can be obtained
for instance by means of optical probe measurements in several points uniformly distributed over
a cross-section of the flow (Bertola, 2002a). Although this technique is time-consuming, because
measurements must cover the whole cross-section with an acceptable spatial resolution, very
often there exist symmetries that allow a considerable reduction of the number of measurements:
for instance, in horizontal pipe flow gravity causes the flow to be symmetrical with respect to the
vertical, so that the void fraction can be measured only over half cross-section. The two-
dimensional polynomial interpolation of the experimental data allows to obtain a surface, which
represents the time average flow structure of the gas-liquid flow.
The cross-sectional average void fraction can be calculated by numerical quadrature of the lo-
cal measurements over the discretized pipe cross-section:
1 N
& ~-""
A~ &G , A ,
(29)
i=l
where A is "the cross-sectional area, Ai the discrete cell area and eGi the local void fraction value
corresponding to the i-th cell.
This technique can be useful to study the flow structure evolution near pipe fittings (such as
contractions, enlargements, bends, valves, orifices, etc.). Figure 6 shows the air-water intermittent
flow structure in several cross-sections along a horizontal pipe with a sharp-edged sudden area
contraction: the void distribution was determined from local void fraction measurements, as de-
scribed above, in seven cross-sections placed 27.5, 15, 1.1 diameters upstream and 1.2, 6, 20,48
diameters downstream of the contraction; the linear interpolation of these distributions allowed
calculating the void distribution over cross-sections placed at different axial positions.
The sudden contraction considerably affects the gas distribution in both the upstream and the
downstream pipe, and its effect grows more and more as the flow approaches the singularity. The
obstacle encountered by the flow causes the level of the incompressible phase in the stratified
region between two slugs to raise, so that the cross-sectional area available for the gas flow gets
smaller; this can be observed for a gas fraction of volume flow up to about 70%, when the gas
flow rate becomes too large to allow an area reduction. In the downstream pipe, the situation is
symmetrical to the upstream pipe, from a qualitative standpoint: the area available on average for
the gas flow is minimum just after the fitting, and grows as the flow goes on. Indeed, a migration
of the air towards the lower part of the cross-section can be noticed.
308 V. Bertola
a)
z/D = -25
d)
z/D =5
b)
z/D = -15 e)
z/0 = 15
c) f)
z/D = -5 z/0 = 25
Figure 20. Axial evolution of the cross-sectional void fraction distribution for an intermittent air-water
flow in a horizontal pipe with abrupt area contraction; images are reconstructed by linear interpolation of
local void fraction measurements performed in seven reference cross-sections.
Far away from the contraction, the void fraction is almost constant in the upper part of the
pipe, and abruptly slopes down to zero in the lower part: this is due to the fact that in this cross-
section plug or slug flow patterns are well established, and no more feel the contraction influence.
The same things can be observed in the cross-section 27.5 diameters upstream. The change in
flow structure along the downstream duct is accompanied by a deceleration of the lower density
phase; hence a void fraction increase and a velocity ratio decrease can be observed.
4 Velocity Measurements
Velocity measurements in two-phase flows are generally much more difficult than in single-phase
flows: in fact, except for the trivial case of homogeneous flow, the phases have different veloci-
ties, so that an ideal instrument should be able to measure both of them. Furthermore, very often
one is not interested in the velocity of one of the phases, but rather in the velocity of some pecu-
liar structure of the flow pattern (for example, a gas bubble or a liquid slug).
The simplest approach suggests to use the same techniques developed for single-phase flow;
these techniques include laser Doppler velocimetry (LDV), hot wire (or hot film) anemometry,
and particle image velocimetry (PIV). Alternatively, methods specifically conceived for two-
phase flows can be used, such as the well known cross-correlation technique or the various and
rapidly developing digital image processing techniques. Among the techniques for single-phase
flows, LDV, which has been popular since the 1970s as a non-intrusive velocimetry technique, is
recognized to give the most reliable and accurate results; this method measures the frequency
shift of a laser beam scattered by a moving object (Doppler effect), and is described in detail, for
instance, by Durst et al. (1981), Heitor et al. (1993), Adrian (1996). Unfortunately, in two phase
flows the laser beam is also scattered by the interfaces between the two fluids, so that the effi-
ciency of LDV is greatly reduced; thus, its applications are limited to dispersed flows with a low
concentration of spherical bubbles or droplets, as summarized by Chigier (1983, 1991 ).
4.1 Velocity in Two-Phase Flows

Let us denote by uk(x) the local, time-average velocity component of the phase kin the mean flow
direction. One can define an average velocity over the cross-sectional area occupied by each
phase:
(30)
If the gas (liquid) velocity is multiplied by the local void (liquid) fraction, one obtains the so-
called superficial velocities:
}a= E:GUG
(31)
jL =E:LUL =(1-&G)uL
The physical meaning of the superficial velocities can be inferred from their average values on
the pipe cross-section:
Jk
AM
I AM
I
=_I_ h(x)dA =_I_ &k(x)uk(x)dA (32)
Thus, superficial velocities indicate the volume flux of each phase; the cross-sectional average
superficial velocity can also be interpreted as the mean velocity each phase would have if the
other phase disappears:
(33)
310 V. Bertola
If one considers the local or the cross-sectional total volume flow rate, the so-called mixture
velocity can be defined:
J=k+iL (34)
J=JG+JL
Sometimes, a coordinate system moving at the mixture velocity can be useful (Zuber and
Findlay, 1965); in this reference frame, the two phases will have a drift velocity:
uDk=uk-j
(35)
Finally, the behavior of two-phase mixtures is often characterized by the velocity ratio, or slip
ratio, which is defined as:
VG
- .
s- uG - AG - VG 1- & (36)
- uL - VL - VL -.,-
AL
Introducing the gas fractions of mass flow, x, and of volume flow, /, the slip ratio can be
written as:
_PLMG1-c_PL x I-c_ .,· 1-c

8- - - .- - - - - - - - - - - - . - - (37)
PGML & PG 1-x & 1-& &
4.2 Cross-Correlation
The cross-correlation R 12 is a measure of the similitude between two signals S 1 and S2. Assuming
that S 1 and S2 can be observed simultaneously as a function of time, we can multiply the values
the two functions have at each instant, and then add together all the results over a given time
interval: thus, if S 1 and S2 are identical the products are always positive and their sum is large,
while if they are different some of the products are positive and some negative, so that the abso-
lute value of the result is smaller. If S 1 and S2 are identical, but S2 is shifted of an unknown
quantity,·, that is, S, = S2(t +,*),we can calculate the sum of the products S2(t)S 2(t + 't) for any
't, and of course the sum will be maximum for 't =,·.From the mathematical standpoint, the defi-
nition of cross-correlation (normalized with respect to time) is:
1 I+T/2
R12 (r)= lim- S 1 (t)S 2 (t+r)dt (38)
T-+too T T/2
If the same signal is contained both in S 1 and in S2, it is strengthened by the cross-correlation,
while any other uncorrelated component is damped; for instance, figure 21 shows the cross-
correlation of two signals containing the same square pulse at different times: R 12 shows a maxi-
mum when 't equals the time lag between the pulses, and rapidly drops to zero for other values: if
the pulses correspond to the passage of an object through two different positions at a known dis-
tance, then the object velocity can be calculated.
2
t
d
1
...
't*
...
I t
d
U=-
't*
Figure 21. Principle of the cross-correlation technique for velocity measurements.
The application of the cross-correlation technique to two-phase flows is quite straightforward

(see e.g. Galaup and Delhaye, 1976), because the passage of the flow structures (e.g. bubbles or
drops in dispersed flow) can be detected by two sensors placed along the direction of the desired
velocity component. The sensors detect some signal in empathy with the flow structures, and can
be based on various techniques, such as ultrasound, optical beams, electrical conductivity or elec-
trical capacitance.
In an ideal case, if a signal detected by the upstream sensor reappears after a certain period t *
at the downstream sensor and the distance d between the two sensors is known, then the velocity
u can be calculated from u = dlt •. In practice, the signals will gradually change while moving
downstream; however, if the position of the downstream sensor is reasonably close to that of the
upstream sensor, the patterns or signals would be sufficiently similar to be recognized by the
correlator, allowing the transit time t to be measured.
Usually, void fraction signals are to be preferred because they are directly related to the flow
structure: in particular, either space-averaged void fractions or the local phase density function
can be used. However, it must be kept in mind that the cross-correlation technique provides only
the most probable propagation velocity of the signal itself, which may have different physical
meanings (or no physical meaning at all) depending on a very large number of parameters.
In general, the signal reflects the passage of gas/liquid interfaces: although the interfaces rep-
resent the boundary of some flow structure (e.g. gas bubbles), their velocity equals that of the
considered flow structure only if the relative (drift) velocity is zero, that is, if the flow structure
appears as "frozen" to a fixed observer. In several circumstances, measurement errors are due to
deformations of the gas/liquid interfaces and/or deviations of the trajectories; while deformations
affect the measurements performed both by integral and by local sensors, trajectory deviations
have worse consequences on local instruments.
31 2 V. Bertola
1I 21
.,
1I 21
Ql ., /- ,
I \I
\- .... '1
I
A B
.-, oL/~\ I
I ~')
~
. . . . ..J
c 0
Figure 22. Examples of possible errors in cross-correlation velocity measurements; (a) deformation of
developing slugs; (b) expanding or collapsing gas bubbles; (c) flattening of the crest of waves on the liquid
surface; (d) trajectory deviation due to turbulent fluctuations of the bubble velocity.
If the flow structure deforms while passing from the upstream to the downstream sensor, the
measured velocity takes into account both the true velocity and the velocity of deformation; on
the other hand, if two local sensors (e.g. optical probes) are used, trajectory deviations might
cause the structures detected by the upstream sensor to pass far away from the downstream one,
and vice-versa. Figure 22 shows four examples among the possible error sources: deformations
can be observed in developing slugs (figure 22a), where the interfaces at the two ends have dif-
ferent velocities, or in expanding/collapsing gas bubbles (figure 22b); deviations can be observed
for instance when the crest of waves in stratified flow flatten (figure 22c ), or when a bubble or a
drop exhibits turbulent velocity fluctuations.
The cross-correlation technique efficiency is also affected by the number of gas-liquid inter-
faces detected, which must be large enough to be statistically significant: in particular, the number
of interfaces is small when the void fraction is close to zero or to one, that is, when one phase is
prevailing on the other. Therefore, one can use the local void fraction as a parameter to prelimi-
nary check the quality of velocity measurements (Bertola, 2002b ).
In order to reduce errors, one must carefully select the rate at which the signals are sampled by
the acquisition system and the distance between the sensors. The sampling rate s should be high
enough to allow the detection of the structures of interest: if L is a characteristic dimension of
these structures (e.g. the mean diameter of bubbles), Nyquist's sampling theorem yields s > 2u/L.
The sensors, as stated above, should be placed at a small distance D from each other to avoid flow
structure changes in between; on the other hand, the distance cannot be too small, because it is
inversely proportional to the velocity measurement uncertainty:
~~u~ = ~u~t~ (39)
where ~t = lis. A thumb rule suggests that the optimal distance between the sensors is 3+5 times
smaller than the characteristic dimension of the flow structures L.
4.3 Hot wire anemometry

The hot wire anemometer, allows one to estimate the velocity U of a fluid from the heat transfer
between a hot wire and the fluid itself. Wires are usually made of platinum or tungsten with di-
ameters of the order of 5 J.!m, and active lengths of the order of I mm. The surrounding fluid can
be considered as a continuum even for very thin wires, since the ratio between the free path of the
fluid molecules and the wire diameter, also known as the Knudsen number, is always very small
(Kn < 0.01 ford> 5 J.!m). The wire temperature is generally not greater than 200°C, so that heat
transfer is essentially due to forced convection and heat conduction through the wire supports,
while free convection and radiation are not significant.
According to Newton's law, the heat flux between the wire and the surrounding fluid is given
by:
(40)
where dis the wire diameter, L the wire length, Tw and Trthe wire and fluid temperatures, respec-
tively. The quantity h [Wm-2K 1] is the so-called heat transfer coefficient, which is a function of
the fluid properties (the thermal conductivity k, the specific heat at constant pressure Cp, the den-
sity p and the viscosity YJ) and of the relative velocity between fluid and wire; this functional
relationship is usually expressed in dimensionless form as:
(41)
where Nu = hdlk, Re = pUd/1], Pr = Cp1]/k, and C, a, p are determined empirically or semi-
empirically. In steady-state conditions, the heat t1ux equals the electric power dissipated in the
wire:
(42)
where I is the electric current and Rw is the wire electrical resistance, which is a function of the
wire temperature, and can be expressed with reference to the resistance at the fluid temperature:
(43)
The combination ofEqs. (41,42,43) allows in principle to find a relationship between the wire
temperature and the fluid velocity; unfortunately, the exact estimation of all the coefficients is
extremely difficult, and leads to inaccurate velocity measurements. Thus, it is preferred to assume
that over a given range of velocities the following expression holds:
314 V. Bertola
continuous
_.----phase
v
~dispersed
phase
Figure 23. Typical voltage output signal of a hot-wire probe in a two-phase dispersed flow; large, U-
shaped fluctuations are related to the passage of the dispersed phase (either bubbles or drops), whereas
smaller fluctuations are related to turbulence in the continuous phase.
2
Rw/ =A+BUn (44)
Rw-Rf
where the coefficients A, Band n are determined from calibration with a flow of known velocity.
Since these coefficients are extremely sensitive to small changes of the wire conditions (e.g. due
to wire fouling), the calibration must be extremely careful, and should be frequently verified. Eq.
(44) suggests that the hot wire anemometer can be used in two ways: the constant-current ane-
mometer, and the constant-temperature anemometer.
In the constant-current mode one measures the change of the wire electrical resistance thanks
to a Wheatstone bridge circuit, whereas in the constant-temperature mode a feedback loop circuit
allows to regulate the current in order to keep the wire temperature (and, consequently, the wire
resistance) steady. The main difference between the two operating modes lies in the dynamic
response, which is much faster for the constant-temperature anemometer. This can be intuitively
understood because the response of the constant current anemometer essentially depends on its
thermal inertia (the wire must change its temperature before a change in the resistance can be
detected), while for the constant temperature anemometer the response depends only on the iner-
tia of the electric circuitry, which is much smaller.
Attempts to use the hot wire anemometer in two-phase flows are dated back to 1960s, as
summarized in the review article by Jones and Delhaye (1976); these works show that the hot
wire anemometer can be used in two-component two-phase flow, whereas in one-component
two-phase flow with phase change the results are not consistent with the calibrated liquid velocity
measurements. Hot wire anemometry is now an established method for the measurements of
volume fraction and statistical velocity parameters of the continuous phase in two-phase dispersed
flows and it can, in principle, be applied to the flow of either a liquid/gas mixture or two immis-
cible liquids (see, for example, Serizawa et al., 1975, Farrar and Bruun, 1989, Liu and Bankoff,
1993a, Al-Deen et al., 1996).
v gas
liquid
t
e =f (Uint' Ap, cr)
A B
Figure 24. Single-probe measurement of the velocity of the dispersed phase; (a) output analog signal,
where the dashed line represents the slope when the gas-liquid interface reaches the probe; (b) deformation
ofthe gas-liquid interface as a bubble touches the hot wire.
When used for two-phase flow velocity measurements, of course, the response of the instru-
ment is different for the two phases. The anemometer output V(t) from a cylindrical hot-film
probe placed in a gas-liquid dispersed flow will be of the form reported in figure 23, as shown by
Farrar and Bruun (1989) and Farrar et al (1995). The signal displays two distinct features: a fluc-
tuating signal caused by the mean velocity and turbulence in the continuous phase interspersed
with aU-shaped transient signal due to the passage of bubbles over the probe. Provided that the
individual bubble influences are correctly eliminated from the whole signal, then the remaining
signal only contains information about the velocity fluctuations in the continuous phase and it is
possible to analyze this part of the signal using statistical techniques to obtain information related
to the corresponding mean velocity and turbulent quantities. In addition to this, to a limited extent
the bubble signals can currently be analyzed to provide statistical information about the bubbles.
Based on existing techniques it is possible to evaluate the volume fraction and the bubble fre-
quency. For instance, Delhaye (1969) proposed to distinguish the liquid and the gas signals in air-
water flow from the probability density function of the output signal amplitude, while Resch and
Leutheusser (1972) identified phase changes in the analog output of the instrument when the
amplitude between two successive extremes was found to exceed a threshold value.
Hamad and Bruun (2000) proposed a signal processing technique that extends the applicabil-
ity of a single hot-film probe to measure the bubble/drop velocity in addition to the continuous
phase velocity. In particular, they studied the impact of a gas bubble (or liquid drop) against the
anemometer wire, and observed that due to surface tension, the gas-liquid interface progressively
wets the wire before the instrument can be immersed in the dispersed phase. As the interface
wraps the wire, the output signal drops (figure 24a); the wrapping angle 8 depends both on the
physical properties of the two fluids and on the interface velocity Uint. as shown in figure 24b. For
small e, the interface velocity is shown to be proportional to d8/dt; on the other hand, the slope of
the output signal is also proportional to d8/dt, so that:
316 V. Bertola
phase interface
,------ -,
: '
!ti
I, _______ J I
l!ft~ I
;
I
laser sheet I _____ __ .J
fi lter
A B
Figure 25. Particle image velocimetry (PIV) setup: a laser sheet illwninates the measurement plane, on
which small tracer particles embedded in the flow are visualized; (a) for single-phase flow, particle images
are recorded on film or ceo camera at right angles to the laser sheet; (b) in two-phase flows, different
seeding is different for the two phases: one kind of seeds simply scatters the light from the laser source,
whereas the other one emits light at a different wavelength, which can be detected by a second camera
equipped with a suitable filter.
Uint = K(dV)
dt A
(45)
where K can be determined from a calibration. However, due to difficulties in measuring the
slope accurately, the uncertainty in this method was estimated to be about ± 10%, which is larger
than for other methods, such as cross-correlation techniques.
4.4 Particle Image Velocimetry (PIV)

Particle image velocimetry (PIV) is a well-known technique that is used for performing whole
field velocity measurements. In the PIV technique, the velocity is measured by recording the
displacement of microscopically small neutral particles, from which the particles velocity can be
inferred.
Figure 25a schematically shows the PIV setup used for velocity field measurements in single-
phase flows: the measurement plane is visualized by a laser sheet that illuminates the flow in the
test section. The flow is seeded with tracer particles, which exactly follow the fluid velocity: in
particular, the tracer particles are considered as ideal when they (1) exactly follow the motion of
the fluid, (2) do not alter the flow or the fluid properties and (3) do not interact with each other.
This kind of seeding is quite different from that used for flow visualization purposes. For flow
visualization, the aim is to make certain flow structures or flow regions visible. This can be ac-
complished by introducing the seeding at a particular location (i.e. inhomogeneous seeding). On
the other hand, a single-exposure PIV recording of an incompressible flow with (ideal) homoge-
neous seeding appears featureless: any flow structure only becomes visible when the velocity
field is evaluated.
real particle image
reflected particle
- ---- -· refracted particle
gas bubble laser sheet
Figure 26. Particle image velocimetry (PIV) in two-phase flows: generation of virtual particle images in
the vicinity of gas-liquid interfaces.
Two or more short laser light pulses with a known time separation illuminate the tracer parti-
cles: thus, particles appear with spacing proportional to the local velocity vector. The image is
captured on film at right angles to the laser sheet, and the velocity field is extracted after image
post-processing of highly resolved PIV photographs (Keane and Adrian 1990; Adrian, 1991).
However, nowadays PIV has developed towards the use of electronic cameras for direct re-
cording of the particle images (see e.g. Willert and Gharib, 1991).
Different tracking methods may be used to process the data. These include techniques such as
auto-correlation, cross-correlation and particle tracking velocimetry (Hassan et al., 1992). In the
autocorrelation technique, post-processing is performed on a single, double exposed image of the
flow; in other words, the same image contains both the particles illuminated by the first laser
pulse and those illuminated by the second one.
The image is divided into small subregions, each containing a number of the seeded particles
(at least 9-10); then, a spatial autocorrelation function (Bendat and Piersol, 1971) of the region
with itself is calculated:
Rll(&,&)= x~oo [ "' dx [F(x,y)F(x+&,y+&)dy (46)

y-+too
where F(x,y) is a scalar property of the image (e.g. luminance). R 11 has a peak in (0,0), because
there is a perfect overlapping of the image with itself, but also a (smaller) peak for (8x*,8y\
where overlapping between the particles illuminated by the two different pulses occurs: thus, 8x •
and 8y* represent the components of the most probable displacement of the particles in the inter-
rogation window, from which the local velocity vector can be calculated if the time separation
between the laser pulses is known. By iterating this procedure over all the subregions, the whole
velocity field is retrieved. In a more recent and popular development, the two images are captured
on separate frames, and a cross-correlation between corresponding subregions is performed.
Autocorrelation and cross-correlation based PIV algorithms have a limited spatial resolution,
which is mainly determined by the size of the interrogation window. This drawback represents an
318 V. Berto1a
important issue especially after the use of digital images has become widespread: as the resolution
and image format of electronic cameras is several orders of magnitude lower than that of a photo-
graphic medium, digitization cannot be ignored (Westerweel 1997). In any situation, the best
spatial resolution can be achieved using particle tracking velocimetry (PTV), which consists of
the determination of the displacements of individual tracers (Stitou and Riethmuller, 2001 ). Re-
cently, new algorithms based on pattern recognition are becoming popular; among them neural
networks (Hassan and Philip, 1997; Cenedese et al., 1992), genetic algorithms and fuzzy logic
techniques seem to have the greatest potential for particle tracking.
The PIV technique can easily be used for multiphase investigations if the different phases can
be distinguished. Several methods are reported to separate the images resulting from different
phases. The use of fluorescent tracers together with a double-camera image acquisition setup is
one way of solving this problem (Philip et al. 1994; Hilgers et al. 1995). With the aid of optical
filters, one may separate the radiation scattered by fluorescent and neutral particles, as shown
schematically in figure 25b: one camera captures both phases, and the other camera with the filter
only captures the seed particle images. Then, by subtracting the images from both cameras, the
two phases can be distinguished.
The use of fluorescent particles allows to suppress the intense reflections near the wall and in-
terface regions and to determine velocity vectors very close to the interface and to the wall. On
the other hand, when cross illuminated by a laser sheet, gas/liquid interfaces originate important
spurious optical effects that must be properly taken into account in order to determine accurately
the surface coordinates from the PIV images (Dias and Riethmuller, 1998; Nogueira et al., 2001).
In particular, these optical effects generate on the PIV image a number of virtual particles, which
have no real counterpart in a corresponding position of the real flow.
As shown in figure 26, the light scattered by particles illuminated by the laser sheet can follow
three different paths before reaching the camera:
- the first path is the direct path that forms the real particle images;
- the second path is due to the a total reflection of light at the gas/liquid interface: it also
reaches the camera and produces spurious particle images seen inside the other phase;
- the third path is due to a double refraction of light beams: if the gas/liquid interface is
curved, the refracted light may reach the interface in a different position, where it if re-
fracted again back into the first medium; these virtual particles are seen inside the second
phase, too.
The result of these optical effects is that it appears essential to use a second method for deter-
mining the gas/liquid interface, since the PIV image itself cannot be trusted.
References
Abuaf, N., Jones, O.C., Zimmer, G.A. (1978), Optical probe for local void fraction and interface velocity
measurements, Rev. Sci. Instrum. 49: 1090-1094.
Adrian, R.J. (1991), Particle-Imaging Techniques for Experimental Fluid Mechanics, Annu. Rev. Fluid
Mech. 22: 261-304.
Adrian, R.J. (1996), Laser Velocimetry. In: R.J. Goldstein (Ed.), Fluid Mechanic Measurements, New
York: Taylor&Francis.
Albouelwafa, M.S.A., Kendall, E.J.M. (1979), Analysis and design of helical capacitance sensors for vol-
ume fraction determination, Rev. Sci. Instrum. 50: 872-878.
Al-Deen, M.F.N., Samways A.L., Bruun, H.H. (1996), Water flow studies using split-film anemometry
Meas. Sci. Techno!. 7: 1529-1535.
Andreussi, P., Di Donfrancesco, A., and Messia, M. (1988), An Impedance Method for the Measurement of
Liquid Hold-Up in Two Phase Flow, Int. J Multiphase Flow, 14: 777-785.
Asali, J.C., Hanratty, T.J. and Andreussi, P. (1985), Interfacial Drag and Film Height for Vertical Annular
Flow, AIChE J 31: 895-902.
Beck M.S., Williams R.A. (1996), Process tomography; a European innovation and its applications, Meas.
Sci. Techno!. 7: 215-224.
Bendat J.S., Piersol A.G. (1971), Random Data: Analysis and Measurement Procedures, New York:
Wiley.
Bertola, V. (2002a), Optical Probe Visualization of Air-Water Flow Structure through a Sudden Area
Contraction, Exp. Fluids 32(4): 481-486.
Bertola, V. (2002b), Slug Velocity Profiles in Horizontal Gas-Liquid Flow, Exp. Fluids 32(6): 722-727.
Bonn, D., Ross, D. (2001), Wetting transitions, Rep. Prog. Phys. 64: 1085-1163.
Brown, R.C., Andreussi, P. and Zanelli, S. (1978), The Use of Wire Probes for the Measurement of Liquid
Thickness in Annular Gas-Liquid Flows, Can. J Chern. Eng. 56: 754-757.
Bruggeman, D.A.G. (1935), Calculation of Different Physical Constants of Heterogeneous Substances,
Ann. Phys. 24: 636-679.
Bruun, H.H. (1995), Hot-wire Anemometry, Oxford: Oxford University Press.
Bruun H H, Samways ALand Ali J (1995), A hot-film study of oiVwater flow in vertical pipes. Proc. 2nd
Int. Conf. on Multiphase Flow (Kyoto, Japan, 3-7 April) paper PI, pp 61-67.
Cartellier, A. (1990), Optical probes for local void fraction measurements: characterization of performance,
Rev. Sci. Instrum. 61: 874-886
Cartellier, A. (1992) Simultaneous void fraction measurement, bubble velocity and size estimate using a
single optical probe in gas-liquid two-phase flows, Rev. Sci. lnstrum. 63: 5442-5453
Cartellier, A., Achard, J.L. (1991), Local phase detection probes in fluid/fluid two-phase flows, Rev. Sci.
Instrum. 62: 279-303.
Cartellier, A., Barrau, E. (1998a), Monofiber optical probes for gas detection and gas velocity measure-
ments: Conical probes, Int. J Multiphase Flow 24(8): 1265-1294.
Cartellier, A., Barrau, E. (1998b), Monofiber optical probes for gas detection and gas velocity measure-
ments: Optimised sensing tips, Int. J Multiphase Flow 24(8): 1295-1315.
Ceccio, S.L., George, D.L. (1996), A Review of Electrical Impedance Techniques for the Measurement of
Multiphase Flow, ASME J Fluid Eng. 118: 391-399.
Censor, Y. ( 1983), Finite series-expansion reconstruction methods, Proc. IEEE 71 (3): 409-419.
Chanson, H. (2002), Air-Water Flow Measurements with Intrusive, Phase-Detection Probes: Can We
Improve Their Interpretation? J Hydr. Eng. 128: 252-255.
Chaucki, J. Laracki, F., Dudukovic, M.P. (1997), Noninvasive tomographic and velocimetric monitoring of
multiphase flows, Ind. Eng. Chern. Res. 36: 4476-4503.
Chigier, N. ( 1983), Drop size and velocity instrumentation, Prog. Energy Combustion Sci. 9: 155-177.
Chigier, N. (1991), Optical imaging of sprays, Prog. Energy Combustion Sci. 17:211-262.
Coney, M.W.E. (1973), The Theory and Application of Conductance Probes for the Measurement of Liq-
uid Film Thickness in Two Phase Flow, J. Phys. E: Scient. Instrum. 6: 903-910.
Costigan, G., Whalley, P.B. (1997), Slug Flow Regime Identification from Dynamic Void Fraction Meas-
urement in Vertical Air-Water Flows, Int. J Multiphase Flow 23: 263-282.
Cartellier, A., Barrau, E. (1998a), Monofiber optical probes for gas detection and gas velocity measure-
ments: Conical probes, Int. J Multiphase Flow 24(8): 1265-1294.
320 V. Bertola
Cartellier, A., Barrau, E. (1998b), Monofiber optical probes for gas detection and gas velocity measure-
ments: Optimised sensing tips, Int. J Multiphase Flow 24(8): 1295-1315.
Cenedese, A., Romano, G.P., Paglialunga, A., Terlizzi, M. (1992), Neural net for trajectories recognition in
a flow, Sixth Int. Symp. on Applications of Laser Techniques to Fluid Mechanics (Lisbon) pp 27.1.1-
27.1.6.
Delhaye, J.M. (1969), Hot-film anemometry, in: Le Tourneau, B.W., Bergles, A.E. (Eds.), Two-Phase
Flow Instrumentation, New York: ASME International.
Devia, F., Fossa, M. (2003), Design and optimisation of impedance probes for void fraction measurements,
Flow Measurement and Instrumentation I4(4-5): 139-149.
Dias, I., Riethmuller, M.L. (1998), PIV in two-phase flows: Simultaneous bubble sizing and liquid velocity
measurements, Proc. 9th Symposium on Laser techniques to fluid mechanics, New York: Springer-
Verlag.
Durst, F., Melling, A., Whitelaw, J.H. (1981), Principles and practice of Laser Doppler Anemometry (2nd
ed.), New York: Academic Press.
Elkow, K.J., Rezkallah, K.S. (1996), Void fraction measurements in gas-liquid flows using capacitance
sensors, Meas. Sci. Techno!. 7: 1153-1163.
Elkow, K.J., Rezkallah, K.S. (1997), Void fraction measurements in gas-liquid flows under 1-g and Jl-g
conditions using capacitance sensors, Int. J Multiphase Flow 23: 815-829.
Farrar, B., Bruun, H.H. (1989), Interaction effects between a cylindrical hot-film anemometer probe and
bubbles in air/water and oil/water flow, J Phys. E: Sci. Instrum. 22: 114-123.
Farrar B., Samways A.L., Ali J., Bruun, H.H. (1995), A computer based hot-film technique for two-phase
flow measurements, Meas. Sci. Techno!. 6: 1528-1537.
Fordham, E.J., Holmes, A., Ramos, R.T., Simonian, S., Huang, S.-M., Lenn, C.P. (1999a), Multi-phase-
fluid discrimination with local fibre-optical probes: I. Liquid/liquid flows, Meas. Sci. Techno!. 10:
1329-1337.
Fordham, E.J., Simonian, S., Ramos, R.T., Holmes, A., Huang, S.-M., Lenn, C.P. (1999b), Multi-phase-
fluid discrimination with local fibre-optical probes: II. Gas/liquid flows, Meas. Sci. Techno!. !0: 1338-
1346.
Fossa, M. (1998), Design and Performance of a Conductance Probe for Measuring the Liquid Fraction in
Two-Phase Gas-Liquid Flows, J Flow Meas. Instrum. 9: 103-109.
Fossa, M., Devia, F. (1999), Theoretical Performance of Impedance Probes for Void Fraction Measure-
ments, Proc. XX International Congress ofRefrigeration, Sydney, Australia.
Galaup, J.P., Delhaye, J.M. (1976), Utilisation des sondes optiques miniatures en ecoulement diphasique
gaz-liquide, application ala mesure du taux de presence local et de vitesse local de la phase gazeuse,
La Houille Blanche I: 17-30.
George, D.L., Ceccio, S.L., O'Hern, T.J., Skollenberg, K.A., Torczynski, J.R. (1998), Advanced material
distribution measurement in mu1tiphase flows: a case study, Proc. ASME-FED 247: 31-42.
Geraets, J.J.M., Borst, J.C. (1988), A capacitance sensor for two-phase void fraction measurement and flow
pattern identification, Int. J Multiphase Flow 14: 305-320.
Gladden, L.F. (1995), Industrial applications ofNMR imaging, Eng. J 56: 149-158.
Gregory, G.A., Mattar, L. (1973), An in-situ volume fraction sensor for two-phase flow of non-electrolytes,
J Canad. Petrol. 12-13: 48-52.
Guenther, R. (1990), Modern Optics, New York: Wiley.
Hamad, F.A., Imberton, F., Bruun, H.H. (1997), An optical probe for measurements in liquid-liquid two-
phase flow, Meas. Sci. Techno/. 8: 1122-1132.
Hamad, F.A., Bruun, H.H. (2000), Evaluation of bubble/drop velocity and slip velocity by a single normal
hot-film probe placed in a two-phase flow, Meas. Sci. Techno!. 11: 11-19.
Hassan, Y.A., Blanchat, T.K., Seeley, C.H. Jr. (1992), PIV flow visualization using particle tracking tech-
niques, Meas. Sci. Techno!. 3: 633--642.
Hassan, Y.A., Philip, O.G. (1997), A new artificial neural network tracking technique for particle image
velocimetry, Exp. Fluids 23: 145-154.
Heerens W.C. ( 1986), Application of capacitance techniques in sensor design, J Phys. E: Sci. Instrum I 9:
897-906.
Heitor M.V., Stamer, S.H., Taylor, A.M.K.P., Whitelaw, J.H. (1993), Velocity, Size, and Turbulent Flux
measurements by Laser Doppler Velocimetry. In: A.M.K.P. Taylor (Ed.), Instrumentation for flows
with combustion, New York: Academic Press.
Herman, G.T. (1980), Image reconstruction from projections - the fundamentals of computerized tomogra-
phy, New York: Academic Press.
Hewitt, G.F. (1978), Measurements ofTwo-Phase Flow Parameters, New York: Academic Press.
Hibiki, T., Mishima, K., Matsubayashi, M. (1995), Application of High-Frame-Rate Neutron Radiography
with a Steady Thermal Neutron Beam to Two-Phase Flow Measurements in a Metallic Rectangular
Duct, Nuclear Technology 110: 422-435.
Hilgers, S., Merzkirch, W., Wagner, T. (1995), PIV measurements in multiphase flow using CCD- and
photo-camera flow visualization and image processing of multiphase flow systems, Proc. 1995
ASME/JSME Fluids Engineering and Laser Anenometry Con[ Exh. FED 209: 151-154.
Hinata, S. (1972), A study on the measurement of the local void fraction by the optical fibre glass probe,
Bull. JSME 15(88): 1228-1235.
Huang, S.M., Stott, A.L., Green, R.G., Beck, M.S. (1988), Electronic transducers for industrial measure-
ment oflow value capacitances, J Phys. E. Sci. Instrum. 21: 242-250.
Jones, O.C., Delhaye, J.M. (1976), Transient and Statistical Measurement Techniques for Two-Phase
Flows: a Critical Review. Int. J Multiphase Flow 3: 89-116.
Kak, A. C., Slaney, M. (1988), Principles of computerized tomography imaging, New York: IEEE Press.
Kang, H.C. Kim, M.H. (1992), The Development of a Flush Wire Probes and Calibration Method for
Measuring Liquid Film Thickness, Int. J Multiphase Flow 18: 423-437.
Karapantios, T.D., Paras, S.V., Karabelas, A.J. (1989), Statistical Characteristics of Free Falling Films at
High Reynolds Numbers, Int. J. Multiphase Flow 15: 1-21.
Kataoka, I., Ishii, M., Serizawa, A. (1986), Local formulation and measurements of interfacial area concen-
tration in two-phase flow, Int. J Multiphase Flow 12(4): 505-529.
Keane R.D., Adrian R.J. (1990) Meas. Sci. Techno!. 1 1202.
Kim, S., Fu, X.Y., Wang, X., Ishii, M. (2000), Development of the miniaturized four-sensor conductivity
probe and the signal processing scheme, Int. J Heat Mass Transfer 43: 4101-4118
Koskie, J.E., Mudawar, I., Tiederman, W.G. (1989), Parallel Wire Probes for Measurements of Thick
Liquid Films, Int. J Multiphase Flow 15: 521-530.
Krane, K.S. (1988), Introductory Nuclear Physics. New York: Wiley.
Kureta, M., Hibiki, T., Mishima, K., Akimoto, H. (1999), Visualization and void fraction measurement of
subcooled boiling water flow in a narrow rectangular channel using high-rate neutron radiography, in:
Two-Phase Flow Modelling and Experimentation 1999 (ed. Celata, G.P., Di Marco, P., Shah, K.):
1509-1514, Pisa: Edizioni ETS.
Le Gall, F., Pascal-Ribot, S., Leblond, J. (200 1), Nuclear magnetic resonance measurements of fluctuations
in air-water two-phase flow: Pipe flow with and without "disturbing" section, Phys. Fluids 13(5):
1118-1129.
Liu T.J., Bankoff S.G. (1993a) Structure of air-water bubbly flow in a vertical pipe-I. Liquid mean veloc-
ity and turbulence measurements, Int. J Heat Mass Transf 36: 1049-1060.
Liu T J and Bankoff S G 1993b Structure of air-water bubbly flow in a vertical pipe-II. Void fraction,
bubble velocity and bubble size distribution Int. J. Heat Mass Transf. 36 1061-1072
Lowe, D., Rezkallah, K.S. (1999), A capacitance sensor for the characterization of microgravity two-phase
liquid-gas flow, Meas. Sci. Techno!. 10: 965-975.
322 V. Bertola
Ma, Y, Chung, N., Pei, B., Lin, W. (1991), Two Simplified Methods to Detennine Void Fractions for Two-
Phase Flow, Nucl. Technology 94: 124-133.
Masuda, Y., Nishikawa, M., Ichijo, B. (1980), New methods of measuring capacitance and resistance of
very high loss materials at high frequancies, IEEE Trans. Instrum. Meas. 29: 28-36.
Maxwell J.C. (1882), A Treatise on Electricity and Magnetism, Oxford: Clarendon Press.
Melnikov, V.I., Kontelev, V.V. (1999), Two-phase flow diagnostic acoustic system based on ultrasound
waveguides, in: Two-Phase Flow Modelling and Experimentation 1999 (ed. Celata, G.P., DiMarco, P.,
Shah, K.): 1515-1519, Pisa: Edizioni ETS.
Melnikov, V.I., Nigmatulin, B.I. (1994), The newest two-phase control devices in LWR equipment based
on ultrasonic and WAT technology, Nucl. Eng. Des. 149: 349-355.
Merilo M., Dechene, R.L., Cicowlas, W.M. (1977), Void fraction measurement with a rotating electric
field conductance gauge, J. Heat Transfer Trans. ASME 99: 330-331.
Mersereau, R.M. (1976), Direct Fourier transform techniques in 3-D image reconstruction, Comput. Bioi.
Med. 6: 247-258.
Mewes, D., Schmitz, D. (1999), Tomographic methods for the analysis of flow patterns in steady and
transient flows, in: Two-Phase Flow Modelling and Experimentation 1999 (ed. Celata, G.P., DiMarco,
P., Shah, K.): 29-42, Pisa: Edizioni ETS.
Mishima, K., Hibiki, T. (1996), Quantitative Method to Measure Void Fraction of Two-Phase Flow Using
Electronic Imaging with Neutrons, Nucl. Sci. Eng. 124: 327-338.
Natterer, F. (1986), The mathematics of computerized tomography, Stuttgart: Teubner Verlag.
Neal, L.G., Bankoff, S.G. (1963), A high resolution resistivity probe for detennination of local void
properties in gas-liquid flow, AIChE J. 9: 490-494.
Nogueira S., Dias, 1., Pinto, A.M.F.R., Riethmuller, M.L. (2001), Liquid PIV measurements around a
single gas slug rising through stagnant liquid in vertical pipes, Proc. lh Int. Conf On Multiphase Flow,
New Orleans.
Okamoto, K., Schmidl, W., Hassan, Y. (1995), New tracking algorithm for particle image velocimetry,
Exp. Fluids 19: 342-347.
Parker, D.S., Hawkesworth, M.R., Broadbent, C.J., Fowles, P., Fryer, T.D., McNeal, P.A. (1994), Indus-
trial positron-based imaging: principles and applications, Nucl Instr. Meth. A349: 583-592.
Philip, O.G., Schmidl, W.D., Hassan, Y.A. (1994), Developments of a high speed particle image veloci-
metry technique using fluorescent tracers to study steam bubbles collapse. Nucl. Eng. Des., 149: 375-
385.
Ramos, R.T., Holmes, A., Wu, X., Dussan, E. (2001), A local optical probe using fluorescence and reflec-
tance for measurement of volume fractions in multi-phase flows, Meas. Sci. Techno/. 12: 871-876.
Reinecke, N., Boddem, M., Petritsch, P., Mewes, D. (1998), Tomographic imaging of the phase distribu-
tion in two-phase slug flow, Int. J. Multiphase Flow 24(4): 617-634.
Resch, F.J., Leuthesser, Leutheusser, J.H. (1972), Le ressaut hydraulique: mesures de turbulence dans la
region diphasique, Houille Blanche 4: 279-293.
Rowland, S.W. (1979), Computer implementation of image reconstruction formulas, in: Image Reconstruc-
tion from Projections Implementation and Applications (ed. Herman, G.T.): 9-80, Berlin: Springer-
Verlag.
Ruder, Z., Hanratty, T.J. (1990), A Definition of Gas-Liquid Plug Flow in Horizontal Pipes, Int. J. Multi-
phase Flow 16: 233-242.
Schlaberg, H.l., Yang, M., Hoyle, B.S. (1996), Real time ultrasonic process tomography for two-
component flows, Electronic Letters 32(17): 1571-1572.
Scott D.M., Williams R.A. (eds) (1995), Frontiers in Industrial Process Tomography, New York: Engi-
neering Foundation.
Sene, K.J. (1984), Aspects ofbubbly two-phase flow, PhD thesis, Trinity College, Cambridge, U.K.
Serizawa, A., Kataoka, 1., Michiyoshi, I. (1975), Turbulence structure of air-bubbly flow II. Local proper-
ties Int. J Multiphase. Flow 2: 235-246.
Stitou, A., Riethmuller, M.L. (2001), Extension ofPIV to super resolution using PTV, Meas. Sci. Techno!.
12: 1398-1403.
Thorn, R., Johansen, G.A., Hammer, E.A. (1999), Three-phase flow measurement in the offshore oil indus-
try - is there a place for process tomography? Proc. 1st World Congress on Industrial Process
Tomography Buxton (UK) pp 228-235.
Tsochatzidis, N.A., Karapantios, T.D., Kostoglou, M.V., Karabelas, A.J. (1992), A Conductance Method
for Measuring Liquid Fraction in Pipes and Packed Beds, Int. J. Multiphase Flow 5: 653-667.
Westerweel, J. (1997), Fundamentals of digital particle image velocimetry, Meas. Sci. Techno!. 8: 1379-
1392.
Willert, C.E., Gharib, M. (1991), Digital Particle Image Velocimetry, Exp. Fluids 10(4): 181-193.
Wu, X., Fordham, E.J., Mullins, O.C., Ramos, R.T. (2000), Single point optical probe for measuring three-
phase characteristics of fluid flow in a hydrocarbon well, USA Patent 6 023 340.
Xie, C.G., Reinecke, N., Beck, M.S., Mewes, D., Williams, R.A. (1995), Electrical tomographic techniques
for process engineering applications, Chern. Eng. J 56: 127-133.
Yang W.Q. (1996), Hardware design of electrical capacitance tomography systems Meas. Sci. Techno!. 7:
225-232.
Zuber, N. Findlay, J.A. (1965), Average volume concentration in two-phase flow systems, Journal of Heat
Transfer (Transactions ASME)C: 453-468.

Berto La 2003

Uploaded by

Copyright:

Available Formats

Berto La 2003

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Berto La 2003

Uploaded by

Copyright:

Available Formats

Two-Phase Flow Measurement Techniques

Laboratoire de Physique Statistique, Ecole Normale Superieure, Paris, France

2 Void Fraction Measurements

V. Bertola (ed.), Modelling and Experimentation in Two-Phase Flow

2.1 Void Fraction Definition

Since two-phase flow is intrinsically inhomogeneous, the value of Eo a priori is a function of

sa{x)=_!_ r Pa(x,t)dt (3)

Figure I. Local void fraction nu tuation s. int gration time.

&c 1(t)=_!_ f Pc(x,t)dl= Lc(t) (4)

Two-dimensional void fraction. The two-dimensional void fraction, or cross-sectional average

&c 2 (t)= _!_

&1 =~ IT{~ 1Pa(x,t)dl} dt = ~ I{~ IT Pa(x,t)dt} dl = L; (7)

&2 =_!_ f {_!_J Pa(x,t)da}dt=_!_J {_!_ f Pa(x,t)dt}da= Aa (8)

&3 =_!_ f {_!_ f Pa(x,t)dv}dt=_!_ f {_!_ f Pa(x,t)dt}dv= Va (9)

2.2 Radiation Attenuation Techniques

Figure 2. Measurement of nc-dimcn ional void fraction by radiation attenuation.

N = N 0 exp(- ~: }xp(- (1 ~: ~ }xp(- ~:) (13)

Figure 4. Schematic of a photomultiplier tube.

2.3 Optical Probes

where k is the propagation vector.

Figure 9. Cleaved optical probe: total internal reflection is never possible.

2.4 Electrical Impedance Methods

G = GLL K(m) (21)

For the helical wound sensor, empirical design criteria are:

.!!... = 1.83 + 3.66 (27)

Figure 17. Feedback loop of iterative reconstruction methods.

Neglecting the conductance of the gas, the measured conductance is:

3.3 Time-Average Flow Visualization

4.1 Velocity in Two-Phase Flows

_PLMG1-c_PL x I-c_ .,· 1-c

Figure 21. Principle of the cross-correlation technique for velocity measurements.

The application of the cross-correlation technique to two-phase flows is quite straightforward

~~u~ = ~u~t~ (39)

4.3 Hot wire anemometry

4.4 Particle Image Velocimetry (PIV)

real particle image

gas bubble laser sheet

Rll(&,&)= x~oo [ "' dx [F(x,y)F(x+&,y+&)dy (46)

You might also like