ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Describe geometrical definitions of void fraction. Describe some measurement techniques. Present homogeneous void fraction model. Describe one-dimensional models. Discuss radial void fraction distributions. Discuss empirical and drift flux type of models. Describe effect of flow pattern on void fraction. Describe LTCM dynamic void fraction measurement technique.

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

1 - Local void fraction

r
1 Pk ( r , t ) = 0

2 - Chordal void fraction

if point r is in phase G if point r is in phase L

local (r, t ) =

1 Pk (r, t )dt t t

L L
L

chordal =

LG LG + LL

3 - Cross-sectional void fraction

AG

4 - Volumetric void fraction

VG

c s =
AL

AG AG + AL

vol
VL

VG = VG + VL

Figure 17.1
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Point-wise void fraction definition:

local = 0 _ or _ 1

1 local (r, t ) = Pk (r, t )dt Time-averaged at a point tt

Chordal void fraction definition: Cross-sectional void fraction definition: Volumetric void fraction definition:
Laboratoire de Transfert de Chaleur et de Masse

LG chordal = L AG c s = A vol VG = V

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

For those new to the idea of a void fraction of a two-phase flow, it is important to distinguish the difference between void fraction of the vapor phase and the thermodynamic vapor quality. To illustrate the difference, consider a closed bottle half full of liquid and the remaining volume occupied by its vapor. The vapor quality is the ratio of the mass of vapor in the bottle to the total mass of liquid plus vapor. If the density ratio of liquid to vapor is 5/1, then the vapor quality is 1/6. Instead, the volumetric void fraction is obtained by applying the expression for vol and in this case would be equal to 1/2. The most widely utilized void fraction definition is the cross-sectional average void fraction, which is based on the relative cross-sectional areas occupied by the respective phases. In this chapter, the cross-sectional void fraction of the gas or vapor phase c-s will henceforth be referred to simply as . Cross-sectional void fractions are usually predicted by one of the following types of methods: Homogeneous model (which assumes the two phases travel at the same velocity); One-dimensional models (which account for differing velocities of the two phases); Drift flux models including radial variations in local void fraction and flow velocity; Models based on the physics of specific flow regimes; Empirical and semi-empirical methods.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Cross-sectional void fraction definition: =

AG A

Mean vapor and liquid velocities in their respective areas:

& & x Q m G uG = = A G

& & 1 x Q m L uL = = A(1 ) L 1

Volumetric flow rate Mass velocity of liquid plus vapor

In homogeneous flow, Vapor uG = uL so that: Rearranging gives [17.2.4]: Quality 1 x G = H = 1 x x Vapor 1 x G Liquid + 1+ density density L G x
L
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Velocity ratio S is often called slip ratio. When uG does not equal uL, we can write their ratio as:

Introducing S into the prior equations for uG and uL , we get: 1 = 1 x G 1+ S x L S > 1 for most flows except some gravity driven down flows when S < 1. When S > 1, the void fraction is smaller than the homogeneous void fraction (which is maximum value).
Laboratoire de Transfert de Chaleur et de Masse

uG S= uL

[17.2.6]

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Utilizing the definition of the velocity ratio and the respective definitions above, a relationship between the cross-sectional void fraction and the volumetric void fraction (the latter obtained by the quick-closing valve measurement technique) can be derived. Returning to the nomenclature used in Section 17.1:

vol =

c s 1 (1 cs ) + cs S

[17.2.7]

Thus, it can be seen that vol is only equal to c-s for the special case of homogeneous flow. For all other cases the velocity ratio must be known in order to convert volumetric void fractions to cross-sectional void fractions.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Momentum flux of a fluid is:

& 2 vH momentum flux = m

Where specific volume of a homogeneous fluid is vH:

v H = v G x + v L (1 x )

For separated flows, the momentum flux is:

2 2 ( 1 x ) vL 2 x vG & momentum flux = m + 1

Differentiating with respect to and setting the momentum flux to zero, velocity ratio is:
Laboratoire de Transfert de Chaleur et de Masse

L S= G

1/ 2

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

This model is based on the premise that the total kinetic energy of the two phases will seek to be a minimum. The kinetic energy of each phase KEk is given by
1 & KE k = k u 2 kQk 2

[17.3.5]

where the volumetric flow rate is in m3/s and uk is the mean velocity in each phase k in m/s. Starting with the definition of the volumetric flow rate for each phase as

& xA m & QG = G

& (1 x )A m & QL = L

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

The total kinetic energy of the flow KE is then

& 2x2 m & xA 1 m & 2 (1 x )2 m & (1 x )A 1 m + L KE = KE k = KE G + KE L = G 2 2 G G L 2 2 (1 )2 2 k =1 L 3 3 3 3 & & ( ) A m x 1 x A m or KE = y = 2 2 + 2 2 2 G (1 ) L 2 3 ( x3 1 x) where the parameter y is: y = 2 2 + G (1 )2 2 L
2

Differentiating parameter y with respect to in the above expression to find the minimum kinetic energy flow gives
dy 2x 3 2(1 x ) =0 = 3 2 + 3 2 d G (1 ) L
3

[17.3.11]

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

The minimum is found when

x L = 1 1 x G
1/ 3

2/3

[17.3.12]

u G L [17.3.13] The velocity ratio S is thus: S = = u L G The velocity ratio is therefore only dependent on the density ratio and the Zivi void fraction expression is

1 1 x G 1+ x L
2/3

[17.3.14]

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

If the fraction of the liquid entrained as droplets in vapor phase is e, summing the kinetic energies of the vapor, liquid in the annular film, and liquid entrained in the vapor (assuming the droplets travel at the same velocity as the vapor), Zivis 2nd 1 method is = [17.3.15]
1 G 2 / 3 1 + e L 1 G 1 G 1 + e + (1 e ) 1 L L 1 + e

1/ 3

Actual value of e is unknown and feasible limits are: For e = 0, the above expression reduces to the prior expression of Zivi for the void fraction, namely [17.3.14]; For e = 1, the expression reduces to the homogeneous void fraction equation, namely [17.2.4].
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Figure 17.2. Influence of entrained liquid fraction on void fraction for ammonia using Zivi (1964) equation [figure taken from Zrcher (2000)].

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Example 17.2: Determine the void fraction for the following vapor qualities (0.01, 0.05, 0.1, 0.25, 0.50, 0.75, 0.95) using the following methods: homogeneous flow, momentum flux model and both of Zivis expressions. The liquid density is 1200 kg m-3 and the gas density is 20 kg m-3. Assume the liquid entrainment is equal to 0.4. Solution: the density ratio of G/L is equal to 0.0167.
Quality, x (1-x)/x =H Homogeneous [17.2.4] [17.3.4] Momentum flux [17.3.14] Zivi #1 [17.3.15] Zivi #2 0.01 99 0.377 0.0726 0.134 0.251 0.05 19 0.759 0.290 0.446 0.665 0.10 9 0.870 0.463 0.630 0.784 0.25 3 0.952 0.721 0.836 0.900 0.50 1 0.984 0.886 0.939 0.960 0.75 0.3333 0.994 0.959 0.979 0.985 0.95 0.0526 0.999 0.993 0.997 0.998

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Smith (1969) assumed a separated flow consisting of a liquid phase and a gas phase with a fraction e of the liquid entrained in the gas as droplets and that the momentum fluxes in the two phases were equal, arriving at the following velocity ratio:
L 1 x + e x G S = e + (1 e ) 1 x 1 + e x
1/ 2

[17.4.1]

For e = 0.4, this gives: =

1
0.78

1 x 1 + 0.79 x

Laboratoire de Transfert de Chaleur et de Masse

G L

0.58

[17.4.2]

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Figure 17.3. Influence of entrained liquid on void fraction by Smith (1969) equation with ammonia [figure taken from Zrcher (2000)].

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

This expression results from simple annular flow theory and application of the homogeneous theory to the fluid density, producing approximately equal frictional pressure gradients in each phase. It is also notable because it goes to the correct thermodynamic limits. Thus, S 1 as x 0, i.e. at very low void fraction the vapor velocity of the very small bubbles should tend towards the liquid velocity since their buoyancy will be negligible. Also, S (L/G)1/2 as x 1, i.e. [17.3.4].

u G L S= = u L H

1/ 2

L = 1 x 1 G

1/ 2

[17.4.3]

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Example 17.3: For the same conditions as in Example 17.2, determine the void fraction for the following vapor qualities (0.01, 0.05, 0.1, 0.25, 0.50, 0.75, 0.95) using the methods of Smith and Chisholm. Assume the liquid entrainment is equal to 0.4. Also, determine the velocity ratio using the Chisholm equation. Solution: the density ratio of G/L is equal to 0.0167.
Quality, x [17.4.2] Smith S [17.4.3] (Chisholm) 0.01 0.274 1.26 0.325 0.05 0.578 1.99 0.614 0.10 0.710 2.63 0.717 0.25 0.852 3.97 0.834 0.50 0.932 5.52 0.916 0.75 0.970 6.73 0.964 0.95 0.993 7.55 0.993

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

The drift flux model was developed principally by Zuber and Findlay (1965), although Wallis (1969) and Ishii (1977) in particular and others have added to its development. Its original derivation was presented in Zuber and Findlay (1965) and a comprehensive treatment of the basic theory supporting the drift flux model can be found in Wallis (1969). Below, methods for determining void fraction based on the drift flux model are presented first for vertical channels and then a method is given for horizontal tubes. Also, the general approach to include the effects of radial void fraction and velocity profiles within the drift flux model is presented.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

The drift flux UGL represents the volumetric rate at which vapor is passing forwards or backwards through a unit plane normal to the channel axis that is itself traveling with the flow at a velocity U where U = UG + UL and thus U remains a local parameter, where superficial velocity of the vapor UG and the superficial velocity of the liquid UL are defined as:

UG = u G
.

[17.4.4a] [17.4.4b]

U L = u L (1 )

Here, uG and uL refer to the actual local velocities of the vapor and liquid and is the local void fraction, as defined by [17.1.1] but dropping the subscript local here. The physical significance of the drift velocity is illustrated in Figure 17.4. These expressions are true for one-dimension flow or at any local point in the flow. Based on these three quantities, the drift velocities can now be defined as UGU = uG U and ULU = uL U.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Figure 17.4

UGU - Vapor velocity relative to the moving reference

Vapor
C0 <U>

Moving reference with velocity C0 <U> in the flow direction

Liquid

Fixed frame of reference

Two-phase flow

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Now proceeding as a one-dimensional flow and taking these parameters as local values in their respective profiles across the channel and denoting the cross-sectional average properties of the flow with < >, which represents the average of a quantity F over the cross-sectional area of the duct as <F> = (AFdA)/A. The mean velocity of the vapor <uG> is thus given by the above expressions to be <uG> = <U> + <UGU> = <UG/>. In addition, <UG> = <uG> and also

& Q < UG > = G A & Q < UL > = L A

uG = < U G > < U > < U GU > = + <> <> <>

[17.4.5a]

[17.4.5b]

The weighed mean velocity G is instead given by G = <uG>/<>. The definition of the drift velocity of the vapor phase UGU yields the following expression for G: [17.4.6]

where <> is the cross-sectional average of the local void fraction.

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

The drift flux is the product of the local void fraction with the local drift velocity [17.4.7a] Using the fact that UGU = uG U combined with the expression for uG in [17.4.4a], rearranging and substituting for UGU in [17.4.7a], the drift flux UGL is also given by the expression:

U GL = U GU

U GL = (1 )U G U L

[17.4.7b]

A distribution parameter Co can now defined as

< U > Co = < >< U >

[17.4.7c]

This ratio accounts for the mathematical difference in averaging and U as a product rather than separately. A weighed mean drift velocity GU can also be defined as

U GU =

< U GL > <>

[17.4.7d]

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Then the following expression is obtained

uG =

< UG > = Co < U > + U GU <>

[17.4.8]

Now dividing through by <U> gives

uG U GU <> = = Co + <U> <> <U>

or

[17.4.9]

< >=

<> U C o + GU <U>

[17.4.10]

where <> is the volumetric quality defined as

& Q < UG > < UG > < >= = = & G& < U > < U G > + < U L > QG + QL
Laboratoire de Transfert de Chaleur et de Masse

[17.4.11]

where the volumetric flow rates of each phase are from [17.3.6] and [17.3.7].

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

For the case where there is no relative motion between the two phases, that is when GU = 0, then

< >=

<> Co

[17.4.12]

Thus, it is evident that Co is an empirical factor that corrects one-dimensional homogeneous flow theory to separated flows to account for the fact that the void concentration and velocity profiles across the channel can vary independently of one another. It follows then for homogeneous flow that [17.4.13] The above expression [17.4.11] demonstrates that < > is the ratio of the volumetric vapor (or gas) flow rate to the total volumetric flow rate. Rearranging that expression in terms of the specific volumes, vG and vL, gives:

< >=< >

xv G < >= xv G + (1 x )v L
Laboratoire de Transfert de Chaleur et de Masse

[17.4.14a]

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

The two-phase density can be expressed as the inverse of the two-phase specific volume v as:

v = xvG + (1 x )v L

1 1 = x 1 x v + G L

[17.4.14b]

Furthermore, the mass velocity can be written as follows:

& =<U> m

[17.4.14c]

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Now, using these three expressions, [17.4.10] for <> can be rewritten as:

x G <>= x 1 x + G L

U GU C + o x 1 x & m + L G
1

[17.4.14d]

Rearranging, the general drift flux void fraction equation becomes

x < >= G

x 1 x U GU C o + + m & L G

[17.4.14e]

The above expression shows that void fraction is a function of mass velocity, while the previously presented analytical theories did not capture this effect. Further note: elsewhere in this book, i.e. other than here in Section 17.4 on drift flux models, the cross-sectional average of the local void fraction <> is written simply as .
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

The drift flux model can be used with or without reference to the particular flow regime as shown by Ishii (1977). The drift flux model, however, is only valuable when the drift velocity is significantly larger than the total volumetric flux, say when GU is larger than 0.05<U>. Also, note that the above drift flux equation [17.4.14e] reduces to the homogeneous void fraction when Co = 1 and either GU = 0 or the mass velocity becomes very large. Several of its applications are discussed below. At elevated pressures, Zuber et al. (1967) have shown that using

C o = 1.13
g ( L G ) U GU = 1.41 2 L
1/ 4

[17.4.15]

[17.4.16]

in [17.4.14e] gives a good representation of their data for R-22 and similar data for water-steam, regardless of the flow regime, and includes surface tension into the method.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

For the bubbly flow regime with one-dimensional vertical upflow of small, isolated bubbles without coalescence, Wallis (1969) has suggested the following equations to use in the drift flux model:

C o = 1 .0

[17.4.17]
1/ 4

g ( L G ) U GU = 1.53 2 [17.4.18] L It is notable that this expression of GU can be interpreted to represent the buoyancy effect of the bubbles on the vapor rise velocity, increasing the vapor velocity with respect to a homogeneous flow. Also for the bubbly vertical upflow regime, Zuber et al. (1967) recommended using [17.4.16] where the value of Co is dependent on the reduced pressure pr and channel internal diameter di, depending on channel size and shape as follows:
For tubes with di > 50 mm: Co = 1 0.5pr (except for pr < 0.5 where Co = 1.2); For tubes with di < 50 mm: Co = 1.2 for pr < 0.5; For tubes with di < 50 mm: Co = 1.2 0.4(pr 0.5) for pr > 0.5; For rectangular channels: Co = 1.4 0.4pr.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

For slug flows, Zuber et al. (1967) recommended using

C o = 1. 2

[17.4.19]
1/ 2

g ( L G )d i U GU = 0.35 L

[17.4.20]

For annular flow, Ishii et al. (1976) proposed using

C o = 1.0
L U L L G U GU = 23 d G i L

[17.4.21] [17.4.22]

The latter expression introduces the effect of liquid dynamic viscosity on void fraction. Ishii (1977) has also given some additional recommendations. For vertical downflow, the sign of GU in [17.4.14e] is changed.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Effect of Non-Uniform Flow Distributions. The definition of Co given by [17.4.7c] may be rewritten for integration of the void fraction profile and the velocity profile as

1 UdA AA Co = 1 1 dA UdA A A A A

[17.4.23]

Its value is thus seen to depend on the distribution of the local void fraction and local phase velocities across the flow channel. As an example, Figures 17.5 and 17.6 depict some experimentally measured values of radial liquid velocity and void fraction profiles for flow of air and water inside a 50 mm bore vertical tube obtained by Malnes (1966). In Figure 17.5, the typical velocity profile for all liquid flow is shown for <>= 0 at two different flow rates.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Figure 17.5. Radial liquid velocity profiles for air-water flow measured by Malnes (1966).

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Figure 17.6. Radial void fraction profiles for air-water flow measured by Malnes (1966).

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Assuming an axially symmetric flow through a vertical circular pipe of internal radius ri and assuming that the flow distributions are given by the following radial functions,

r U = 1 Uc r i

[17.4.24]

w c w

r = 1 r i

[17.4.25]

where the subscripts c and w refer to the values at the centerline and the wall, Zuber and Findlay (1965) integrated [17.4.23] to obtain the following expression for the distribution parameter Co

Co = 1 +

w 2 1 m+n+2 < >

[17.4.26]

when expressed in terms of w or

m+2 c n Co = 1+ m + n + 2 < > m + 2

when expressed in terms of c.
Laboratoire de Transfert de Chaleur et de Masse

[17.4.27]

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

They noted that if the void fraction is uniform across the channel, i.e. if w = c = <>, then it follows that Co = 1. If c > w, then Co > 1. On the other hand, if c < w, then Co < 1. Furthermore, if m is assumed to be equal to n and the flow is adiabatic (w = 0), then [17.4.26] reduces to

n+2 Co = n +1

[17.4.28]

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Lahey (1974) presented an interesting picture of the variation in radial void fraction and values of Co for different types of flow patterns in vertical upflow. Figure 17.7. Diabatic void fraction profiles for selected flow regimes as presented by Lahey (1974).

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Rouhani and Axelsson (1970) correlated the drift velocity for vertical channels as

g(L G ) U GU = 1.18 2 L
where

1/ 4

[17.4.29]

Co = 1.1 for mass velocities greater than 200 kg m-2 s-1; Co = 1.54 for mass velocities less than 200 kg m-2 s-1. Instead, the following correlation of Rouhani (1969) can be used for Co over a wide range of mass velocities:

gd i C o = 1 + 0.2(1 x ) 2 m &

2 L

1/ 4

[17.4.30]

This expression is valid for void fractions larger than 0.1. By combining these expressions with [17.4.14e], it is possible to obtain an explicit value for <>.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Example 17.4: Determine the local void fractions using the RouhaniAxelsson expression [17.4.30] for the following qualities (0.1, 0.50, 0.95) for a fluid flowing at a rate of 0.1 kg/s in a vertical tube of 22 mm internal diameter. The fluid has the following physical properties: liquid density is 1200 kg m-3, gas density is 20 kg m-3 and surface tension is 0.012 N m-1. Solution: The mass flux for this situation is 263.1 kg m-2 s-1 while the gravitational acceleration g = 9.81 m s-2.
Quality, x Co [17.4.30] GU [17.4.29] <> [17.4.14e] 0.10 1.262 0.10525 0.653 0.50 1.146 0.05847 0.852 0.95 1.015 0.00585 0.984

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Horizontal tubes. All the above methods are for vertical tubes. For horizontal tubes, Steiner (1993) reports that the following method of Rouhani (1969) is in good agreement with experimental data, whose modified form was chosen in order to go to the correct limit of <> = 1 at x = 1:

Co = 1 + c o (1 x )

[17.4.31]

where co = 0.12 and the term (1-x) has been added to the other expression to give:

g( L G ) U GU = 1.18(1 x ) 2 L

1/ 4

[17.4.32]

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

R410A, T = 40C 1 0.9 0.8

Fig.17.8. Comparison of three methods for R-410A in 8 mm tube showing effect of mass velocity in Steiner/RouhaniAxelsson equation.

0.7 Void Fraction 0.6 0.5 G = 500 kg/m 2s 0.4 0.3 Homogeneous 0.2 0.1 0 0 0.2 0.4 0.6 0.8 1 Vapor Quality Zivi Rouhani G = 75 kg/m 2s G = 200 kg/m 2s

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Wojtan, Ursenbacher and Thome (2003) have measured 238 time-averaged cross-sectional void fractions in an 13.6 mm horizontal glass tube using a new optical measurement technique, processing about 227,000 images. Figure 17.9 shows a comparison using [17.4.31] and [17.4.32] and the homogeneous model.

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Butterworth (1975). Many void fraction methods have a standard form of:

1 x = 1 + n B x

n1

G L

n2

L G

n3

[17.5.1]

which allows for a comparison to be made and also inconsistencies to be noted, e.g. if the densities and viscosities in both phases are equal, i.e. L = G and L = G, then the void fraction should be equal to the quality x. This means that n3 and n1 should be equal to 1.0 in all void fraction equations in order to be valid at the limit when the critical point is approached. The exponent on the density ratio is always less than that of the homogeneous model. Importantly, some methods include the influence of viscosity while others do not.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Which is the best general approach? The drift flux model appears to be the preferred choice for the following reasons: 1. Plotting measured flow data in the format of G versus <U> yields linear representations of the data for a particular type of flow pattern. This was known empirically to be the case, and then Zuber and Findley arrived at the reason for this linear relationship analytically with [17.4.8]. Hence, analysis of experimental data can yield values of Co (the slope) and GU (the yintercept) for different types of flow patterns using [17.4.8]. 2. Various profiles can be assumed for integrating [17.4.23] and hence appropriate values of Co can be derived for different flow patterns and specific operating conditions (e.g. high reduced pressures). 3. As a general approach, the drift flux model provides a unified approach. It is capable of being applied to all directions of flow (upward, downward, horizontal and inclined) and potentially to all the types of flow patterns. 4. Reviewing of the expressions for GU, the drift flux model includes the important effects of mass velocity, viscosity, surface tension and channel size.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Plotting measured flow data in the format of G versus <U>, as shown in Figure 17.10, yields linear (or nearly linear) representations of the data for a particular type of flow pattern. This was known empirically to be the case, and then Zuber and Findley arrived at the reason for this linear relationship analytically with [17.4.8]. Hence, analysis of experimental data can yield values of Co (the slope) and GU (the y-intercept) for different types of flow patterns using [17.4.8]. Note that homogeneous flow gives the diagonal straight line dividing adiabatic upflow from adiabatic downflow.

uG =

< UG > = C o < U > + U GU <>

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Dynamic Flow Visualization and Image Processing of TwoPhase Flows New Experimental Technique of LTCM Image Acquisition with Laser

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Dynamic Flow Visualization and Image Processing of TwoPhase Flows New Experimental Technique of LTCM Image Processing Technique to Measure Flow Phenomena
Am

m =
m

Ai Am

Ai

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Dynamic Flow Visualization and Image Processing Applied to Two-Phase Flows New Experimental Technique of LTCM

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

It is the cross-sectional void fraction, not the volumetric void fraction, that is required for prediction of flow boiling coefficients, flow pattern transitions and two-phase pressure drops. Cross-sectional void fractions cannot be determined using the quick-closing valve method since that gives volumetric void fractions, except for homogeneous flow. The relationship between the cross-sectional void fraction and the volumetric void fraction Vol is Cross-

Vol =

S is the velocity or slip ratio between the two phases. Since typically S > 1, incorrectly using the volumetric void fraction Vol in place of the cross-sectional void fraction overpredicts the latters value, and in some cases gives void fractions larger than the homogeneous void fraction (for which S = 1)!!! For example, if = 0.20 and S = 1.5, then Vol = 0.273, or a 36% increase in value! Hence, the assumption that the quick-closing valve technique can yield a cross-sectional void fraction database (done in recent published studies) is quickly shown to be fundamentally incorrect and gives unreasonable values.
Laboratoire de Transfert de Chaleur et de Masse

1 (1 ) + S

sectional void fraction

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Void fractions in two-phase flows over tube bundles are much more difficult to measure than for internal channel flows. Mass velocities of industrial interest tend to be lower than for internal flows. For evaporation and condensation in refrigeration systems: 5 to 40 kg m-2 s-1. In partial evaporators and condensers common to the chemical processing industry with single segmental baffles, design range is from 25 to 150 kg m-2 s-1. For vertical two-phase flows across tube bundles the frictional pressure drop tends to be small compared to the static head of the two-phase fluid at low Gs. The void fraction thus becomes the most important parameter for evaluating the two-phase pressure drop via the local two-phase density of shell-side flows. Even though shell-side void fractions have been studied less than internal channel flows, they are still important for obtaining accurate thermal designs. In flooded type evaporators with close temperature approaches in refrigeration and heat pumps, the effect of the two-phase pressure drop on the saturation temperature may be significant in evaluating the log mean temperature difference.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

The earliest study was apparently that of Kondo and Nakajima (1980) who investigated air-water mixtures in vertical upflow for staggered tube layouts. They utilized quick-closing valves at the inlet and outlet of the bundle that included not only the tube bundle but also the entrance and exit zones. Based on the mass of liquid in the bundle, the void fraction was determined. Their tests covered qualities from 0.005 to 0.90 with mass velocities of 10 to 60 kg m-2 s-1, where the mass velocity as standard practice is evaluated at the minimum cross section of the bundle similar to single-phase flow. They found that the void fraction increased with superficial gas velocity while the superficial liquid velocity had almost no affect on void fraction. The number of tube rows had an effect on void fraction, probably resulting from inclusion of their inlet and exit zones when measuring void fractions. The quick-closing valve technique yields volumetric void fractions, which are only equal to cross-sectional void fractions when S = 1; when S >1, the volumetric void fraction is larger than the cross-sectional void fraction.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Shrage et al. (1988) made similar tests with air-water mixtures on inline tube bundles with a tube pitch to diameter ratio of 1.3, again using the quickclosing valve technique, ran tests for qualities up to 0.65, pressures up to 0.3 MPa and mass velocities from 54 to 683 kg m-2 s-1. At a fixed quality x, the void fraction was found to increase with increasing mass velocity. They offered a dimensional empirical relation for predicting void fractions by incorporating a multiplier to the homogeneous void fraction, referred to as H. A non-dimensional version was obtained using further refrigerant R-113 data and it is as follows:

ln x vol = 1 + 0.123 0.191 H FrL

where the liquid Froude number was defined as

[17.6.1]

& m FrL = 1/ 2 L (gd o )

Laboratoire de Transfert de Chaleur et de Masse

[17.6.2]

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Ishihara et al. (1980) presented the following method for predicting the local void fraction in tube bundles based on the two-phase friction multiplier of the liquid:

1 = 1 L

[17.6.3]

Their liquid two-phase friction multiplier L is given by

8 1 L = 1 + + 2 X tt X tt

[17.6.4]

Xtt is the Martinelli parameter for both phases in turbulent flow across the bundle, which reduces to the simplified form

1 x G X tt = x L

0.57

L G

0.11

[17.6.5]

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Fair and Klip (1983) offered a method for upflow on horizontal reboilers:

(1 )2 =

1 L

[17.6.6]

Their liquid two-phase friction multiplier L is given by

L = 1 +

20 1 + 2 X tt X tt

[17.6.7]

where Xtt is for both fluids turbulent is the same as in Ishihara et al. above.

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Feenstra, Weaver and Judd (2000) developed a completely empirical expression to predict the velocity ratio S, where the expression for obeys the correct limits at vapor qualities of 0 and 1. From dimensional analysis, the following parameters influenced S: two-phase density, liquid-vapor density difference, pitch flow velocity of the fluid, dynamic viscosity of the liquid, surface tension, gravitational acceleration, the gap between neighboring tubes, tube diameter, tube pitch and the frictional pressure gradient. The two-phase density and density difference were included as they are always key parameters in void fraction models. The tube pitch Ltp and tube diameter D were included for their influence on the frictional pressure drop. Furthermore, the surface tension was selected since it affects the bubble size and shape and the liquid dynamic viscosity L was included because of its affect on bubble rise velocities.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Feenstra et al. proposed the following void fraction prediction method:

1 = (1 x ) G 1 + S x L
where the velocity (or slip) ratio S is calculated as:

[17.6.8]

S = 1 + 25.7 (Ri Cap )

0.5

(L

tp

D)

[17.6.9]

In this expression, Ltp/D is the tube pitch ratio and the Richardson number Ri is defined as:

Ri =

( L G )2 g (L tp D )
&2 m
[17.6.10]

The Richardson number represents a ratio between the buoyancy force and the inertia force. The mass velocity, as in all these methods for tube bundles, is based on the minimum cross-sectional flow area like in single-phase flows.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Feenstra et al., cont.: The characteristic dimension is the gap between the tubes, which is equal to the tube pitch less the tube diameter, Ltp-D. The Capillary number Cap is:

Lu G Cap =
& xm uG = G

[17.6.11]

It represents the ratio between the viscous force and the surface tension force. The mean vapor phase velocity uG is determined as

[17.6.12]

Hence, an iterative procedure is required to determine the void fraction using this method.

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Feenstra et al., cont.: This method was successfully compared to air-water, R-11, R-113 and water-steam void fraction data obtained from different sources, including the data of Schrage, Hsu and Jensen (1988). It was developed from triangular and square tube pitch data with tube pitch to tube diameter ratios from 1.3 to 1.75 for arrays with from 28 to 121 tubes and tube diameters from 6.35 to 19.05 mm. Their method was also found to be the best for predicting static pressure drops at low mass flow rates for an 8-tube row high bundle under evaporating conditions (where the accelerational and frictional pressure drops were relatively small) by Consolini, Robinson and Thome (2006). Hence, the Feenstra-Weaver-Judd method is thought to be the most accurate and reliable available for predicting void fractions in vertical twophase flows on tube bundles.

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Example 17.5: Determine the void fraction and velocity ratio using the Feenstra-WeaverJudd method at x = 0.2 for R-134a at a saturation temperature of 4C (3.377 bar) for a mass velocity of 30 kg/m2s, a tube diameter of 19.05 mm (3/4 in.) and tube pitch of 23.8125 mm (15/16 in.). The properties required are: L = 1281 kg/m3; G = 16.56 kg/m3; = 0.011 N/m; L = 0.0002576 Ns/m2. Solution: Begin by assuming a value for the void fraction, where here the value of 0.5 is taken as the starting point in the iterative procedure. First, the mean vapor velocity is determined to be:

0.2(30) uG = = 0.725 m / s 0.5(16.56)

Next the Capillary number is determined:

Cap =

0.0002576(0.725) = 0.0170 0.011

The Richardson number Ri is determined next:

2 ( 1281 16.56) (9.81)(0.0238125 0.01905) Ri = = 83.0

302

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Example 17.5, cont.: The void fraction is then:

1 = 0.432 ( ) 1 0 . 2 16 . 56 1 + 25.4 0.2 1281

The second iteration begins using this value to determine the new mean vapor velocity and so on until the calculation converges. After 6 iterations, becomes 0.409 (with an error of less than 0.001).

Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

Grant and Chisholm (1979), using air-water mixtures, tested a baffled heat exchanger with 4 crossflow zones and 3 window areas (note: a window refers to where the flow goes through the baffle cut and is hence longitudinal to the tubes rather than across the tubes). They studied stratified flows and measured the liquid levels in the first and fourth baffle compartments, noting that the void fraction was less in the first baffle compartment with respect to the fourth compartment for all mass velocities tested. They proposed:
1 = 1 L x 1+ 1 x G 1 K 2

[17.6.13]

The empirical factor K2 is a velocity ratio evaluated as

where m is obtained by fitting the method to experimental data. They noted that this method worked well at low qualities but overestimated the measured void fractions at higher qualities.
Laboratoire de Transfert de Chaleur et de Masse

L K2 = G

m 2 m

L G

1 m 2m

[17.6.14]

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

17.1: Determine the void fraction for the following vapor qualities (0.01, 0.1 and 0.25) using the following methods: homogeneous flow, momentum flux model and both of Zivis expressions. The liquid density is 1200 kg/m3 and the gas density is 200 kg/m3. Assume the liquid entrainment is equal to 0.4. 17.2: Determine the void fraction for a vapor quality of 0.1 using the second of Zivis expressions using the properties above. Assume liquid entrainments equal to 0.0, 0.2, 0.4, 0.6, 0.8 and 1.0. 17.3: Determine the local void fractions using the Rouhani expressions [17.4.31] and [17.4.32] for the following qualities (0.1, 0.50, 0.95) for a fluid flowing at rates of 0.05 and 0.2 kg/s in a horizontal tube of 22 mm internal diameter for the same conditions as in Example 17.4. Compare and comment. The surface tension is 0.012 N/m. 17.4: Determine the local void fraction using the drift flux model assuming bubbly flow for the following qualities (0.01, 0.05, 0.1) for a fluid flowing at a rate of 0.5 kg/s in a vertical tube of 40 mm internal diameter. The fluid has the following physical properties: liquid density is 1200 kg/m3, gas density is 20 kg/m3 and surface tension is 0.012 N/m.
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

17.5: Determine the local void fraction using the drift flux model assuming slug flow for the following qualities (0.05, 0.1) for the same conditions as in Problem 17.4. 17.6: Determine the local void fraction using the drift flux model assuming annular flow at a quality of 0.5 for a fluid flowing at a rate of 0.5 kg/s in a vertical tube of 40 mm internal diameter. The fluid has the following physical properties: liquid density is 1000 kg/m3, gas density is 50 kg/m3, surface tension is 0.05 N/m and the liquid dynamic viscosity is 0.0006 Ns/m2. 17.7: Determine the local void fractions using the method of Ishihara et al. for the following qualities (0.05, 0.1, 0.50) at a mass velocity of 100 kg/m2s flowing across a tube bundle with tubes of 25.4 mm outside diameter. The fluid has the following physical properties: liquid density is 1200 kg/m3, gas density is 20 kg/m3, surface tension is 0.012 N/m and the liquid and vapor dynamic viscosities are 0.0003 Ns/m2 and 0.00001 Ns/m2
Laboratoire de Transfert de Chaleur et de Masse

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

17.8: Determine the local void fractions using the method of Fair and Klip for the same conditions as in Problem 17.7. 17.9: Derive expression [17.3.13] from [17.3.12]. 17.10: Derive Zivis second void fraction expression with entrained liquid in the gas phase, i.e. expression [17.3.15]. 17.11: Derive expressions [17.4.26] and [17.4.27]. 17.12: Prove expression [17.2.7].

Laboratoire de Transfert de Chaleur et de Masse