Professional Documents
Culture Documents
www.elsevier.com/locate/ces
Received 8 August 2003; received in revised form 23 June 2004; accepted 18 July 2004
Available online 11 September 2004
Abstract
A model for bubble formation from a single submerged orifice is developed using the boundary-integral method. Since the flow field
is assumed to be irrotational, potential-flow theory is used to predict the growth of the bubble. The effects of the surface tension and
the liquid circulation around the bubble are included in the calculation. Predictions of bubble shape, chamber pressure and the effect of
surface tension are presented to compare with reported experimental data.
䉷 2004 Elsevier Ltd. All rights reserved.
Keywords: Bubble formation; Boundary-integral method; Mathematical modeling; Orifice; Potential-flow; Thermodynamics
Over the past decades, numerous theoretical and exper- The system under consideration is shown in Fig. 1. Gas
imental studies have been reported in the field of bubble is pumped into a chamber at a constant flow rate, Q, and
formation. In addition, many models have been developed bubbles out through a small orifice into a bath of liquid. The
to predict bubble formation, such as those by Marmur and viscosity of the liquid is assumed negligible and the flow
Rubin (1976), Pinczewski (1981), Tsuge and Hibino (1983), is assumed to be irrotational. Therefore, a velocity potential
Tan and Harris (1986) and Hooper (1986). exists, u = ∇ , with Laplace’s equation
In the model proposed by Hooper (1986), the boundary-
integral method, a popular and successful numerical tech- ∇ 2 = 0. (1)
nique, was used to predict bubble formation. The boundary-
integral method is based on Green’s formula that enables
Then Bernoulli integral is applied between the liquid side
us to reformulate the potential problem as the solution of a
of the bubble surface and a point in the liquid chosen at a
Fredholm integral equation. This formulation has the effect
large distance from the orifice on the plane z = 0,
of reducing the dimension of the problem by one, which has
the advantage of rendering the technique computationally
efficient, yet rigorous. Some applications of this method can * 1 2 Pl Po
+ |u| + + gz = (2)
be found in Blake et al. (1986, 1987), Bonnet (1995) and *t 2 l l
Power (1995).
where Po = P∞ + l gH is the hydrostatic pressure at the
orifice, P∞ is the system pressure and H is the total height
of liquid above the orifice.
∗ Corresponding author. Tel.: +65-6874-6360; fax: +65-6779-1936. It is convenient to express the dynamic boundary con-
E-mail address: chetanbh@nus.edu.sg (R.B.H. Tan). dition in terms of a material derivative by employing the
0009-2509/$ - see front matter 䉷 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2004.07.064
180 Z. Xiao, R.B.H. Tan / Chemical Engineering Science 60 (2005) 179 – 186
According to the Young–Laplace equation, the local curva- In order to solve the problem of the two-dimensional
ture of certain point on the surface is defined as bubble computationally, firstly the surface is represented by
n points which divide the surface into (n−1) elements. Then
1 1
= + (15) an isoparametric linear approximation is used for the surface
R1 R2 and the functions. Thus on each segment, Sj (j =1, L, n−1),
where R1 and R2 are principal radii of curvature, on vertical we have
and horizontal planes, respectively. From analytical geom-
r() = rj (1 − ) + rj +1
etry, the principal radii of curvature may be represented by
the following equations (Gray, 1998): z() = zj (1 − ) + zj +1
() = j (1 − ) + j +1
1 z
1 z
denote the first and second where the parameter is in the range (0, 1).
derivatives with respect to r which can be obtained through Collocation points are chosen to be these n points on the
curve fitting written in the form z = z(r) along the bubble surface, yielding n equations with the following form:
surface.
n−1
* 1
2i + [(1 − )j + j +1 ] ds
2.3. Volumetric growth rate of bubble j =1 Sj
*n |pi − qj |
n−1
*j *j +1
With the values of the normal velocity *
*n
at a number of = (1 − ) +
1
ds
points on bubble surface Sb , the bubble volumetric growth Sj * n *n |pi − qj |
j =1
rate can be obtained by the integral (i = 1, L, n) (21)
dVb * where pi is the ith point chosen as collocation point
= 2 r ds. (17)
dt Sb *n with cylindrical coordinates pi = (ri , zi , 0) and qj is
any point on the segment Sj with cylindrical coordinates
2.4. Normal velocity qj = (r(), z(), ).
After assembly, the set of equations has the following
For any sufficiently smooth function , which satis- matrix structure:
fies Laplace’s equation in a domain having a piecewise
smooth surface S, Green’s integral formula can be written *
A = B (22)
as (Brebbia, 1978; Jaswon and Symm, 1977): *n
* where A and B are n × n matrices with the following values
c(p)(p) + (q) G(p, q) dS of items aij and bij
s *n
*
= ((q))G(p, q) dS (18) * 1
s *n aij =2
ij + (1 − ) ds
Sj *n |pi − qj |
where p ∈ +S, q ∈ S, **n is the normal derivative outward * 1
from S, G(p, q) is Green’s function + ds (23)
Sj −1 *n |pi − qj −1 |
1
G(p, q) = (19) 1 1
|p − q| bij = (1−) ds+ ds (24)
Sj |pi −qj | Sj −1 |pi −qj −1 |
and
*
4, if p ∈ and are vectors as follows:
c(p) = . *n
2, if p ∈ S
= [1 , 2 , LL, n ]T
2.5. Axisymmetric bubbles
T
* *1 *2 *n
As the bubble is assumed to be axisymmetric, a cylin- = , , LL, .
*n *n *n *n
drical coordinate system is used. Two-dimensional integrals
are reduced further to one dimension by integrating through The integrals are performed analytically through the
the polar angle. In this case the Green’s functions involve azimuthal angle from 0 to 2 to yield complete ellip-
complete elliptic integrals. tic integrals and then numerically evaluated by using the
182 Z. Xiao, R.B.H. Tan / Chemical Engineering Science 60 (2005) 179 – 186
= q'(ri, -zi)
0 [(r() + ri ) + (z() − zi ) ]
2 2 1/2
2
d
× 1/2 (25)
4r( )r 2
i cos 2
0
1−
(r()+ri ) +(z()−zi )
2 2
dz dr 2ri dz where q
is the image of q in the z = 0 plane.
× (r() + ri ) − (z() − zi ) − 2
d d k () d
E(k) 2ri dz 2.7. Tangential velocity with cubic spline interpolation
× + K(k) (28)
1 − k 2 () k 2 () d
To complete the specification of the surface velocity at a
given point, the tangential component must be approximated
where E(k)= 02 (1−k 2 sin2 )1/2 d is the complete elliptic from the given values of the velocity potential (pi ) where
integral of the second kind. pi ∈ Sb . We choose the point at the top of the bubble on the
Z. Xiao, R.B.H. Tan / Chemical Engineering Science 60 (2005) 179 – 186 183
2.9. Time stepping (A4b) and (C1) of Hooper, 1986) have been amended in our
present work.
To update the position of the bubble and the potential at
its surface through time, an iterative trapezium rule, Euler’s
Method, is used 4. Result and discussion
ri (
+
) =
ri (
) + ui,r (
)
The model is verified through comparison between theo-
zi (
+
) =
z i (
) +
ui,z (
)
retical computation and experimental data for low viscosity
D i (
) liquids as reported in the literature.
i (
+
) = i (
) +
(35) For the air-water system of Kupferberg and Jameson
D
(1969) with the following experimental conditions: gas flow
where ui,r and
ui,z are radial velocity and axial velocity at rate Q = 16.7 cm3 /s; the radius of the orifice Ro = 0.16 cm;
ith point of the bubble surface respectively. chamber volume Vc = 2250 cm3 ; height of the liquid
The procedure is summarized in the following steps: H = 15.24 cm, the instantaneous shapes and bubble growth
curves and chamber pressure fluctuation are shown in
(i) Initialize all variables. Fig. 4. The bubble shapes obtained by the present model
(ii) Increment time by t. are shown in Fig. 4(a), which agree approximately with the
(iii) The field equation (1) is solved using boundary-integral experimental shapes shown in Fig. 4(b) obtained from the
method. Using the Green’s formula approach the nor- original photographs in Kupferberg and Jameson (1969).
mal velocity is found directly by solving Eq. (22) Fig. 4(c) compares the simulated bubble growth curve and
numerically. The tangential velocity is solved by cubic chamber pressure fluctuation with the experimental data.
spline interpolation. For the CO2 –water system of LaNauze and Harris (1974)
(iv) Use the velocities found in (iii) to update position of at elevated pressure with following experimental conditions:
the surfaces. gas flow rate Q = 10 cm3 /s; the radius of the orifice Ro =
(v) Use the dynamic condition, Eq. (34), to update the sur- 0.16 cm; chamber volume Vc = 375 cm3 ; height of the liquid
face potentials. H = 10 cm; system pressure P∞ = 0.69 MN/m2 , the instan-
(vi) Go back to (ii) and repeat until detachment happens. taneous shapes and bubble volume are shown in Fig. 5. The
bubble shapes obtained by the present model are shown in
Fig. 5(a). Fig. 5(b) compares the simulated bubble growth
curves with the experimental data.
3. Improvements over Hooper’s (1986) model LaNauze and Harris (1974) reported that they observed
“double bubbling” under these conditions. The experimen-
The proposed model is generally based on Hooper’s tal growth curve in Fig. 5(b) shows two distinct periods
(1986) approach to modeling bubble formation using the of growth. This phenomenon can be attributed to the wake
boundary-integral method. However, we have made several effect of preceding bubbles (Zhang and Tan, 2000). Our
significant improvements and developments in our model. present model does not account for the wake effect, since
Firstly, the selection of dimensionless numbers represent- the liquid is assumed to be quiescent initially. Nevertheless,
ing the physical and dimensional parameters allows the nat- the predicted final bubble volume at detachment agrees well
ural and a priori formulation of the dimensionless equation with the experimental data, as shown in Fig. 5(b).
of motion (Eq. (34)), instead of relying on the questionable Surface tension is one of the contributing factors influ-
iterative method to define length and time scales proposed encing the bubble volume. To verify the effect of surface
by Hooper (1986). tension, data are collected for liquids of different surface
Secondly, Hooper (1986) does not explicitly account for tensions, water ( = 72.7 dyn/cm) and petroleum ether
the curvature of the bubble surface in relating Pl to Pb . In- ( = 27.1 dyn/cm), using the same orifice (Ro = 0.2 cm) by
stead, it was merely stated that due to numerical instabilities, Davidson and Schüler (1960). Fig. 6 shows the variation
the surface tension number was always set to zero, thereby of bubble volume at detachment with gas flow rate for dif-
ignoring the effect of altogether and rendering Pl equal to ferent liquids (Davidson and Schüler, 1960, as reported by
Pb in all the solutions. Ramakrishnan et al., 1969) and the calculated results are
Thirdly, the use of a realistic boundary condition (Eq. compared with the experimental data. It can been seen that
(31)) at the point where the bubble surface meets the orifice at lower gas flow rates, surface tension has little effect on
plate allows us to relate the bubble shape to the dynamic the bubble volume, while at higher flow rates, the lowering
three-phase (i.e. gas–liquid–solid) contact angle, , instead of surface tension produces smaller bubbles. It can be noted
of the arbitrary shape criterion introduced in Hooper’s (1986) that our model predictions match the experimental trends
work. closely.
Furthermore, several material errors in the theoretical de- These figures indicate that the results computed by present
velopment of the earlier model (see Eqs. (31), (A3a), (A4a), model are in good agreement with the experimental data.
Z. Xiao, R.B.H. Tan / Chemical Engineering Science 60 (2005) 179 – 186 185
1.8 1
0.9 t=37.5ms
1.6 t=30ms
0.8
t=49.5ms
1.4 0.7
t=20ms
t=40ms 0.6
z (cm)
1.2
0.5
t=30ms t=10ms
z (cm)
1 0.4
0.3 t=5ms
0.8 t=20ms
0.2 t=0
0.6 0.1
t=10ms
0.4 t=5ms (a) -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
r (cm)
0.2 t=0
0.1
t=30ms
0.8
t=20ms 0.0
0.6 0 10 20 30 40
t=10ms (b) Time (ms)
0.4
t=5ms
t=0 Fig. 5. (a) Computed sequence of bubble shapes during formation from
0.2
LaNauze and Harris (1974). (b) Bubble growth curve from LaNauze and
0.0 Harris (1974).
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
(b)
r (cm)
2.0
Experimental Volume Experimental Pressure Model Predictions
Gas/Liquid σ Ro Exp. Cal.
2.5
(dyn/cm) (cm)
100 Air/Water 72.7 0.20
Air/Petroleum 27.1 0.20
2.0 1.5
Bubble Volume Vb (cm )
Ether
3
Bubble volume (cm3)
80
Pc -Po (N/m )
2
1.5
60
1.0
1.0
40
20 0.5
0.5
0 0.0
0 10 20 30 40 50 60
(c) Time (ms)
0.0
0 5 10 15 20 25 30 35 40
Fig. 4. (a) Computed sequence of bubble shapes during formation from Volume Flow Rate Q (cm3/s)
Kupferberg and Jameson (1969). (b) Approximate experimental shapes
from Kupferberg and Jameson (1969). (c) Bubble growth curve and Fig. 6. Effect of surface tension on bubble size in inviscid liquid from
chamber pressure fluctuation from Kupferberg and Jameson (1969). Davidson and Schüler (1960).
186 Z. Xiao, R.B.H. Tan / Chemical Engineering Science 60 (2005) 179 – 186
b bubble
c gas chamber