Professional Documents
Culture Documents
This study reports an experimental investigation into the hydrodynamic behavior of an external-
loop airlift reactor (ALR) for the air-water system. Three distinct flow regimes are
identifiedsnamely homogeneous, transition, and heterogeneous regimes. The transition between
homogeneous and heterogeneous flow is observed to occur over a wide range rather than being
merely a single point as has been previously reported in the literature. A gas holdup correlation
is developed for each flow regime. The correlations fit the experimental gas holdup data with
very good accuracy (within (5%). It would appear, therefore, that a deterministic equation to
describe each flow regime is likely to exist in ALRs. This equation is a function of the reactor
geometry and the system’s physical properties. New data concerning the axial variation of gas
holdup is reported in which a minimum value is observed. This phenomenon is discussed and
an explanation offered. Discrimination between two sound theoretical modelssnamely model
I (Chisti et al., 1988) and model II (Garcia Calvo, 1989)sshows that model I predicts satisfactorily
the liquid circulation velocity with an error of less than (10%. The good predictive features of
model I may be due to the fact that it allows for a significant energy dissipation by wakes behind
bubbles. Model I is now further improved by the new gas holdup correlations which are derived
for the three different flow regimes.
{ }{ }
An important aspect in the modeling of the hydrody-
namics of airlift reactors is the relationship between the Rate of energy Rate of energy dissipation
dependent variables of gas holdup and liquid circulation input due to due to internal turbulence
rate and the independent variables of superficial gas ) +
isothermal gas and friction between
velocity (independently controllable), the physical prop- expansion the gas-liquid interface
{ }
erties of the fluids, and reactor geometry. This last
variable has a strong influence on the hydrodynamics Rate of energy losses due
and makes it difficult to compare the results from to friction between the (1)
different sources. fluids and the reactor
Several empirical correlations are reported in the
literature (Chisti, 1989), but these do not give any or
∑Ë + ∑Ê
fundamental understanding about the hydrodynamics
and they are thus of limited use for extrapolation. Ein ) (2)
Notable theoretical models are those of Chisti et al.
(1988), Garcia Calvo (1989), and Garcia Calvo et al. ∑Ë ) Ër + Ëd + Ët + Ëb (3)
(1991) which are based on a view of the energy balance
and those of Hsu and Dudukovic (1980), Merchuk and ∑Ê ) Êr + Êd + Êt + Êb (4)
* Author to whom all correspondence should be addressed. where r, d, t, and b refer to riser, downcomer, top, and
† E-mail: mabashar@pixie.udw.ac.za.
bottom sections of the reactor. Because of the negligible
S0888-5885(97)00461-2 CCC: $15.00 © 1998 American Chemical Society
Published on Web 02/14/1998
1252 Ind. Eng. Chem. Res., Vol. 37, No. 4, 1998
drift of gas with respect to the liquid in both the top and changes in flow direction) are given by
and bottom sections, the terms Ët and Ëb are essentially
zero. VLr2
The energy input due to the isothermal expansion of Êt ) m̆LrKt (14)
an ideal gas (supplied at a pressure of Ph and a 2
superficial gas velocity of Ugr) is given by the following
(Chisti, 1989): VLd2
Êb ) m̆LdKb (15)
2
(
Ein ) UgrArPh ln 1 +
Ph)
FLgh
(5) where V is the linear velocity and K is the friction loss
coefficient. The relations between the linear velocities
of the liquid in the riser and the downcomer and the
2.1. Model I (Chisti et al., 1988). The model of superficial liquid velocity in the riser are obtained as
Chisti et al. makes the following assumptions: follows:
(1) Steady-state conditions. (2) Isothermal conditions.
(3) Negligible mass transfer between the gas and the ULr
liquid. (4) The energy losses terms due to the skin VLr ) (16)
friction in the riser and the downcomer negligible (Êr (1 - ro)
()
≈ 0, Êd ≈ 0) in comparison to the other dissipation
terms. This assumption is justified by the experimental ULd Ar ULr
VLd ) ) (17)
evidence of Lee et al. (1986) for low-viscosity Newtonian (1 - do) Ad (1 - do)
fluids (e.g., water). (5) The pressure drop due to
acceleration negligible (Wallis, 1969). Substitution of eqs 14-17 in eq 4 gives
The energies associated with turbulence and internal
[ ]
friction, Ër and Ëd are obtained by an energy balance
Kt Kb(Ar/Ad)2 ULr2
on the riser and on the downcomer. The energy balance
for the riser is given by the following:
∑ Ê ) m̆Lr
(1 - ro) 2
+
(1 - )2 2
(18)
do
Ein + pressure energy loss ) Substitution of eq 13 for ΣË and eq 18 for ΣÊ in eq 2
potential energy gain + Ër (6) gives the superficial liquid velocity in the riser as
[ ]
or (neglecting the mass of the gas compared to that of 2gh(ro - do) 0.5
ULr ) (19)
the liquid): Kt Kb(Ar/Ad)2
+
Ein + m̆Lr(1 - ro)gh ) m̆Lrgh + Ër (7) (1 - ro)2 (1 - do)2
(1 - ro) ( )(
Ar Ugr
Ad ro )
- Vs (22)
VLd2 ULd3
ΣÊ ) Kf m̆Ld ) KfFLAd )
2 2
[ ( )(
Ar Ugr
)]
3
1
KfFLAd (1 - ro) - Vs (23)
2 Ad ro
(
UgrPh ln 1 +
FLgh
Ph )
- VsroFLgh -
( )[
Ar 2
(
Ugr
)]
3
1
KfFL (1 - ro) - Vs ) 0 (24)
2 Ad ro
3. Experimental Section Figure 1. Experimental setup for the external loop airlift
reactor: (1) air compressor; (2) pressure regulator and filter; (3)
The pilot plant external loop airlift reactor used in turbine flowmeter; (4) rotameters; (5) sparger; (6) water drainage;
this work is shown in Figure 1. The reactor was made (7) pressure tapping; (8) electromagnetic flowmeter; (9) water inlet;
of borosilicate glass with an approximated working (10) inverted U-tube manometer; (11) disengagement tank.
volume of 0.725 m3. The major dimensions are sum- Table 1. Major Reactor Dimensions
marized in Table 1. Air was introduced through a
circular perforated plate sparger containing 193 holes diameter height dispersion height
(m) (m) (m)
of 1-mm diameter on a 11-mm square pitch. The
sparger was designed according to the criteria given by riser 0.225 6.75 6.06
Ruff et al. (1978) and was located about 0.94 m from downcomer 0.225 6.75
disengager 1.58 × 0.38 × 0.50
the base of the reactor. The flow rate of the air was
measured by a turbine flowmeter and controlled by a
needle valve just downstream of it. A set of rotameters operated at room temperature and atmospheric pres-
was included for visual indications only. The liquid sure. The dispersion height was kept at 6.06 m.
circulation velocity in the downcomer was measured by
an electromagnetic flowmeter. The pressure drop along 4. Results and Discussion
the riser was measured by a set of inverted U-tube
manometers connected to four taps 1.1 m apart in three 4.1. Characterization of Flow Regimes. The
sections of the riser. The first tap was located at 1.36 characterization of various flow regimes has been
m above the sparger which is greater than the maxi- described by Shah et al. (1982) for bubble columns. The
mum distance necessary for the equilibrium bubble size, bubbly (homogeneous) flow regime is characterized by
5 times the column diameter (Joshi et al., 1990). The almost uniform sized bubbles with equal radial distri-
pressure measurement system enabled the calculation bution. The heterogeneous (churn turbulent) flow re-
of the local gas holdup as well as the overall gas holdup gime is characterized by large bubbles moving with high
for individual and a combination of sections. A single rise velocities in the presence of small bubbles. Het-
inverted U-tube manometer was used to measure the erogeneous flow is typically characterized by a nonuni-
gas holdup in the downcomer. For all experiments air form radial gas holdup profile. The third regime (slug
and filtered tap water were used and the reactor was flow) occurs only in small diameter columns (up to 0.15
1254 Ind. Eng. Chem. Res., Vol. 37, No. 4, 1998
Figure 4. Linear gas velocity vs the total flux in the riser. Figure 5. Local gas holdup profiles in the riser for various
superficial gas velocities.
Table 2. Parameters for the Drift Flux Model
drift
distribution velocity
system dr (m) coefficient (m/s) ref
air-water 0.225 1.07 0.538 this work
air-water 0.200 1.20 0.260 Verlaan (1987)
air-water 0.240 1.13 0.280 Nicol and Davidson (1988)
air-water 0.140 1.03 0.330 Merchuk and Stein (1981)
Figure 7. Parity plot for the overall gas holdup in the riser. Figure 8. Gas holdup in the downcomer vs the overall gas holdup
in the riser.
The Maclaurin expansion of the term 1/(1 - roC1) is
holdup in the riser is shown in Figure 8. The following
1 correlation is obtained:
) 1 + roC1 + (roC1)2 + (roC1)3 + ... (28)
(1 - roC1)
do ) 0.174ro 0.007 e ro e 0.06 (35)
Definitely this expansion is singular at roC1 ) 1.
A similar linear relationship with a slope of 0.79 and
Assuming a power law dependence as shown in eq 29,
intercept of -0.057 was reported by Bello et al. (1985).
The differences between eq 35 and the correlation of
ULr ) aUbgr (29) Bello et al. (1985) is due to the fact that the overall gas
holdup in the downcomer depends on the geometry and
then eq 27 becomes the efficiency of the disengagement tank.
4.3. Models Application. These gas holdup cor-
Ugr ) (C3Ubgr + C2)(ro + ro2C1 + ro(roC1)2 + relations developed in this study for the three flow
regimes are now implemented, mainly in model I.
ro(roC1)3 + ...) (30) 4.3.1. Model I (Chisti et al., 1988). The reactor
geometry affects the values of Kb and Kt. In this study
Neglecting the high-order terms (since ro is typically the friction coefficients at the bottom and the top of the
rather small) and also the term roC2 gives reactor are estimated using the engineering correlations
as Kb ≈ Kt ) 1.8 (Streeter and Wylie, 1979). Different
ro ) RUβgr (31) values of Kb are reported in the literature, but this value
(1.8) is almost the same as the value obtained by
The coefficients (R, β) are obtained for all flow regimes Verlaan (1987). Figure 9a shows a comparison between
by analyzing our experimental data using nonlinear experimental and predicted values of the superficial
regression analysis. A correlation for each regime is liquid velocity in the riser using model I (eq 19). The
obtained as follows: developed correlations (eqs 32-34) for the gas holdup
(a) homogeneous regime are implemented in this model. It is clearly shown in
Figure 9a that this model predicts the superficial liquid
ro ) 0.29U0.74 0 e Ugr e 0.02 (32) velocity in the riser with satisfactory accuracy and the
gr
difference between experimental and predicted values
is lower than (10%. Figure 9b shows the prediction of
(b) transition regime
model I when the well-known gas holdup correlation of
Hills (1976) is used for the different flow regimes.
ro ) 0.37U0.81
gr 0.02 < Ugr e 0.043 (33)
Ugr
(c) heterogeneous regime ro ) ULr > 0.3 m/s
0.24 + 1.35(Ugr + ULr)0.93
(36)
ro ) 1.58U1.232
gr Ugr > 0.043 (34)
Substitution of this correlation in eq 19 allows the
Figure 7 shows the predicted values of the overall gas calculation of the superficial liquid velocity in the riser
holdup from these correlations vs the experimental (ULr) for each value of the superficial gas velocity in the
values. The power law correlations are in good agree- riser (Ugr). The resulting nonlinear algebraic equation
ment with the experimental values ((5%). The plot of is solved numerically by an IMSL (International Math-
the gas holdup in the downcomer vs the overall gas ematics and Statistics Library) subroutine called ZSPOW
Ind. Eng. Chem. Res., Vol. 37, No. 4, 1998 1257
Figure 9. Parity plot for the superficial liquid velocity in the riser
(model I): (a) using different gas holdup correlations for various Figure 10. (a) The overall gas holdup in the riser vs the square
flow regimes from this work and (b) using gas holdup correlation superficial liquid velocity in the downcomer and (b) parity plot
of Hills (1976) for various flow regimes. for the superficial liquid velocity in the riser (model II).
based on a variation of Newton’s method which uses a square of the superficial liquid velocity in the down-
finite difference approximation to the Jacobian and comer. The best fit was made by using the method of
takes precautions to avoid large step sizes or increasing least squares. From the slope of the line the total
residuals. It is obvious that the deviation of the model friction coefficient (Kf) was found according to eq 37 to
prediction becomes significant ((21%) as shown in be equal to 4.43. From this figure, as far as the linear
Figure 9b and the overall predictiveness is unacceptable. relationship exists, the total friction coefficient is inde-
The prediction of model I is very sensitive to the value pendent of the liquid velocity and the gas holdup. A
of ro. comparison between the total friction coefficient (Kf) and
4.3.2. Model II (Garcia Calvo, 1989). Hsu and the total value of Kb and Kt (Kb + Kt ) 3.6) shows that
Dudokovic (1980) developed a relation between gas most of the energy losses due to the friction between
holdup and the liquid velocity based on momentum the fluids and the reactor are dissipated at the top and
balance in which the pressure difference between the the bottom of the reactor.
riser and the downcomer is considered to be due to the The slip velocity was calculated according to Har-
friction losses in the reactor. mathy’s equation (1960):
[σLg(FL - Fg)
]
1/4
Kf
ro ) U 2 (37) Vs ) 1.53 (38)
2gh Ld FL2
This relation has been used by many workers (Verlaan, where the surface tension of the water (σL) is 0.075 N/m.
1987; Merchuk and Stein, 1981). Figure 10a shows the Equation 24 is a nonlinear algebraic equation. Know-
experimental measured gas holdup in the riser vs the ing Kf and Vs this equation is solved by ZSPOW to
1258 Ind. Eng. Chem. Res., Vol. 37, No. 4, 1998
predict ro for each value of the superficial gas velocity j ) mean gas holdup, dimensionless
(Ugr) and the corresponding superficial liquid velocity F ) density, kg/m3
is obtained from eq 25. Figure 10b shows that the σ ) surface tension, N/m
difference between the experimental and the predicted
Abbreviations
values of the superficial liquid velocity in the riser is
less than (16%. It is obvious that model I is more ALR ) airlift reactor
accurate than model II; this may be due to consideration Subscripts
of energy dissipated by the wakes behind the bubbles
and then may be considered different for very different b ) bottom of the reactor
flow regimes. Despite the lower accuracy of model II d ) downcomer
in this study, it still has an attractive feature of g ) gas
predicting the liquid superficial velocity and the gas L ) liquid
holdup on the basis of the total friction coefficient (Kf), o ) overall
the slip velocity (Vs), and the superficial gas velocity r ) riser
(Ugr). s ) slip
t ) top of the reactor
Shah, Y. T.; Kelkar, B. G.; Godbole, S. P. Design Parameters Young, M. A; Carbonell, R. G.; Ollis, D. F. Analysis of Local Two-
Estimations for Bubble Column Reactors. AIChE J. 1982, 28 Phase Hydrodynamics. AIChE J. 1991, 37 (3), 403.
(3), 353. Zuber, N.; Findlay, J. A. Average Volumetric Concentration in
Streeter, V. L.; Wylie, E. B. Fluid Mechanics; McGraw Hill: New Two-Phase Flow Systems. J. Heat Transfer 1965, 87, 453.
York, 1979.
Verlaan, P. Modeling and Characterization of an Air-lift-Loop Received for review July 2, 1997
Bioreactor, Ph.D. Thesis, Agricultural University at Wagenin- Revised manuscript received November 14, 1997
gen, Wageningen, The Netherlands, 1987. Accepted November 15, 1997
Wallis, G. B. One-Dimensional Two-Phase Flow; McGraw-Hill:
New York, 1969. IE9704612