A General Design Equation for Air Lift Pumps

Operating in Slug Flow

Air lift pumps are finding increasing use where pump reliability and low mainte- N. N. CLARK
nance are required, and where corrosive, abrasive, or radioactivefluids must be han- Particle Analysis Center
dled. Although air lifts are used in nuclear fuel reprocessing plants, no general, theo- West Virglnia University
retically sound equation has been proposed in the literature for tall air lift design. Morgantown, WV 26506
Such an equation is developed from two-phase flow theory to predict the height to
which an air lift pump operating in the slug flow regime can lift a given volumetric and
flow rate of liquid, given the air flow rate and pressure at the point of gas i n t d u c - R. J. DABOLT
tion. The widely used drift-flux model for the prediction of holdup is combined with
an appmximate relationshipto predict pmsure loss, and is substituted into the total Allied General Nuclear Services
pressure differential. Integration of the resulting equation provides an explicit for- Barnwell, SC
mula for the calculation of lift. Ekperimental work using a variety of liquids in a 38
mm dia. air lift test installation supports the new design equation and demonstrates
its flexibility.

SCOPE
Air lift pumps provide a reliable means of raising corrosive or tions, although valid for the test data used, are not necessarily ~ p -
abrasive liquids and slurriesbecause they incorporateno moving plicable over the very wide range of operatingconditions and lift
parb to erode and wear. Such pumps are ideal also for handling heights (from 2 to 2,000 m) typical in air lift applications. Nicklii
highly radioactiveliquids because they r e q h virtuallyno main- (1963)provided a theoreticallysound analysis of air lift pumps us-
tenance and can be treated as remote units. Although air lift ing a momentum balance. However, this analysis was accurate
pumps have a wide variety of possible applications, most studies only in the design of short pumps, since there was no provision for
have been concerned with dewatering mines or raising oil from variation in gas volumetric flow rate over the tube length. For
dead wells. More recently, the importance of air lifts in moving taller pumps, the method has to be applied incrementally. In the
liquids at nuclear fuel reprocessing plants has been realized, so analysis below, a differential momentum balance is integrated
that more accurate design equations are required. To date, liquid over the whole pump length to provide a closed-form equation
flow rate in the air lift has been predicted using either an energy that is valid for air lifts of any height operating in bubble or slug
balance or empirical correlations. The energy balance method, two-phase flow. This new equation compares favorably with data
although valid, cannot take into account losses in the pump ex- in the literature, and with experimental data from a 38 mm dia.
cept in terms of an overall efficiency, which is generally not testinstallation.
known and cannot be predicted accurately. Empirical correla-

CONCLUSIONS AND SIGNIFICANCE

Results from a 38 mm air lift test installation support a new de- equation is able to predict the performance of very tall air lifts
sign equation, based on a two-phase flow momentum balance. used by Shaw (1920) in mine dewatering, and so appears valid
The test apparatus was constructedto provide for theoperation of over a wide range of operating conditionsand configurations. In
the air lift in either vacuum or overpressure conditions, and four using this equation, one need know only the air flow rate, essen-
differentliquids were used in the tests.Operating curves plotted tial measurements of the pump, and friction factor for the flow in
from the results over a range of air flow rates agree with curves order to predict the liquid flow rate. Since all of these terms are
given by the new design equation. In addition, the new design readily available to the design engineer, this new equation will
prove easier to use than energy balances, which require a knowl-
Correspondence concerningthis paper should be addressed to N. N. Clark. edge of the pump efficiency, or empirical correlations,which are
R. J. Dabolt is presently with Chem Nuclear Systems, Columbia, SC. often cumbersome and not universally applicable.

Page 56 January, 1986 AlChE Journal (Vol. 32, No. 1)

 n
INTRODUCTION GAS OUT

Air lift pumps are used to raise liquids or slurries from wells or
vessels, particularly where submerged mechanical pumps are un-
desirable or where simplicity of construction is required. In the LIQUID O U l
simplest case, a vertical tube is partially submerged in liquid in a
vessel, and air is introduced at or near the bottom of the tube. Be-
ing less dense than the liquid, the buoyant air-liquid mixture
formed in the air lift tube rises and is expelled at the top of the
pump, Figure la. An alternative U-tube airlift arrangement, in
which the submergence is provided by a downcomer leg, is illus-
trated in Figure lb. The two configurations are equivalent, ex-
cept that some significant energy losses may occur in the U-tube
arrangement due to the flow of fluid in the downcomer.
I n theory, air lift pumps couldoperate with any interdispersion
of the air and liquid phases, but in practice most air lift pumps op-
erate in the slug flow regime, which persists Over a wide range of
air and liquid velocities (Taitelet al., 1980; Clark, 1984a). In slug
flow large bubbles of air aresurrounded by an annular film of liq-
uid in contact with the pipe wall. These air bubbles are separated
by slugs of liquid (spanning the whole tube diameter) which may
contain a few smaller air bubbles. Although sufficient literature
exists to describe the slug flow regime (Govier and Aziz, 1972;
Griffith and Wallis, 196l), current design techniques are still Flgure lb. Air llff pump: U-tube arrangement,
based on semiempirical equations such as those presented by
Shaw (1920) and O’Neill(l975). These equations become inaccu-
rate when applied to any nonstandard air lift designs. In this pa- (at atmospheric pressure), V,, required to raise some volume of
per, a new design equation is derived from two-phase flow theory water, V,, through L feet, is given by
and is shown to be applicable over a wide range of operating con- V,/Vu; = e&./[P,ln(P,/PJ]
ditions and pump heights.
where Pois the pressure at the bottom of the tube, where gas was
LiteratureSurvey introduced, and P2is the pressure at the discharge, typically at-
mospheric pressure. However,in the air lift pump there are en-
Although air lifts have been used in mine dewatering since the ergy loses that can be ascribed to losses in the bulk flow and to the
eighteenth century, until the last three decades two-phase flow slippageof the gas bubbles through the liquid. Due to these losses,
theory was inchoate and could not be applied to pump design. most early air lifts operated at 35 to 55 % energy efficiency, so that
The earliest theoretical approaches relied on an energy balance, the theoretical energy balance could not be applied except with
equating the energy possessed by the compressed air introduced an energy efficiencyterm, typically 50 % (Shaw, 1920). Modified
at the base of the pump to the useful work done in raising the liq- energy balance equations in various forms have appeared in the
uid from the level in the well to the top of the pump. Shaw (1920) literature in the intervening time, and are still recommended for
noted and derived several such analyses. Assuming isothermal op- air lift design today (O’Neill, 1975). Nevertheless, the energy bal-
eration, for a pump with 100% efficiency the volume of free air ance is insufficient for accurate design, since the true pump effi-
ciency is seldom known until the pump has been constructed, or
the design subjected to more rigorous two-phase flow analysis. In
cases where low submergences or narrow bore pipes are used,
these energy balances overpredict flow rate to a significant de-
gree.
In the petroleum engineeringfield, design of gas lifts for recov-
ering oil from slow or dead oil wells has followed a different path,
and most designs rely on graphical correlations, or empirical
equations based on extensive full-scale well data (Lawson and
Brill, 1975; Brown, 1967; Ros, 1961).Whereas most chemical en-
gineering air lift designs are concerned with raising water or
aqueous solutions, the petroleum engineer must deal with a range
of viscosities and densitiesin considering the oil to be lifted. In ad-
dition, the oil under pressure at the base of the well generally con-
tains varying amounts of light fractions, which become gas at
higher elevations(lower pressures) in the tube. Thereforethe mass
flow of gas in the string is not constant and further complicates
the prediction of gas lift performance. Much effort in the presen-
tation of petroleum engineering correlationshas been devoted to
dealing with these problems, but the work is unsuited to the de-
sign of air lifts in the chemical and nuclear engineeringfields.
Nicklin (1963)provided the first useful analysis of the air lift
pump based on a momentum balance. The pressure gradient in
the air lift tube was separated into the hydrostatic head term and
frictional losses. The hydrostatic head was predicted from the
Figure la. Alr lift pump: alr iffl tube immersedIn a well or vessel. holdup of liquid and gas, which were evaluated using a two-

AlChE Journal (Vol. 32, No. 1) January, 1986 Page 57

phase flow drift-flux model, which is discussed in more detail in therefore the pressure at the base of the air lift tube, at the point of
the analysis below. From an energy balance point of view, losses air introduction. In the case where the submergence is provided
due to the bubble slippage are accounted for in the prediction of by a downcomer, as in a U-tube air lift (Figure lb), Pomust be re-
holdup, and frictional losses are accounted directly, using a modi- duced by the losses occurring in the downcomer tube up to the
fied single-phase flow (D’Arcy) equation. However, Nicklin’s point of gas introduction. The need to account for these losses has
differential analysis was really valid only at a single height in the been emphasizedby Dabolt and Plummer (1980)and Husain and
pump, since the equation assumed a constant value for the gas Spedding (1976).
superficial velocity, which increases with increasing pump The top of the air lift tube, height xl, will discharge to atmo-
height. Stenning and Martin (1968) successfully applied with sphere, or into a vessel at a known pressure, Pz.Acceleration ef-
Nicklin analysis to a short test pump, using an average air superfi- fects due to the expansion of the gas over the pump height are a
cial velocity, but this approach would overpredict the required very small fraction, typically less than 1 % ,of the overall gradient
air rate in taller pumps (Dabolt and Clark, 1985). in the tube and can be neglected. Therefore, where the small con-
More recently, Husain (1975) and Husain and Spedding (1976) tribution due to the air phase density in the air lift tube is also dis-
proposed an original energy balance, assuming a spectrum of en- regarded, the total differential for the pressure gradient in the air
ergies for the bubbles in the air lift tube. The air sparger was lift tube is given by
viewed as an energy-emitting orifice, and the analysis assumed - dP = [elg(l - E) + F]dn (2)
inviscid liquid and gas. However, this analysisstillrequired a term
for “efficiencyof gas usage.” Moreover, the definition of this effi- where t is the cros-sectional average air void fraction, and F is the
ciency differed from that used in the early energy balances, being irreversible loss per unit pipe length, at a height x in the pipe. By
inversely proportional to the gas fraction present. Jeelani et al. integrating Eq. 2 between the entrance condition (x = 0, P = Po),
(1979)have demonstrated that the “constants”derived in the Hu- and the exit condition at the top of the air lift (x = x,, P = P2), it is
sain and Spedding analysis may vary when applied to narrow possible to obtain an explicit formula to predict the total height of
bore lifts, so that the analysis would not appear to be suited to the pump, r2, and hence the lift, 22 - 21. However, in order to in-
general air lift design. Although this analysis supplies the correct tegrate, both the air void fraction and the frictional pressure loss
shape for most of the air lift operating curve, the constants for the must be predicted as a function of air and liquid flow rates. Sten-
equation must be found by regression. ning and Martin (1968) chose to evaluate the void fraction, E , at
Dabolt and Plummer (1980) examined a wide range of data for the mean pressure in the pump, but this approach may be inaccu-
air lifts in the 12 to 50 mm &a. range, for application in nuclear rate in the case of tall pumps, where E may vary greatly over the
fuel reprocessing. The submergence ratio, or ratio of submer- tube height.
gence to total pump height, was found to be the dominant varia- In vertical slug flow one may not simply equate the in situ air
ble governing lift performance, and the following design equa- void fraction, E , with the volumetric flowing air fraction, W,/
tion was proposed (W, + W,), because the velocities of the air and liquid phases dif-
fer in the pipe. First, the slugs of air rise relative to the liquid, so
V, = ln(1 - V,/A)/B that liquid flows back in the annular film surrounding the air
where V, is the volume of air required to lift a volume V, of water, slug. Second, in occupying the central region of the pipe, the slug
and where A and B are quadratic functions of the submergence is traveling in a region of flow which is moving faster than the
ratio, with coefficients dependent on pipe diameter. Frictional cross-sectionalaveragevelocity Both of these factors contribute to
losses before the sparger, and in the air lift tube, and pressures in raising the slip ratio, or ratio of air to liquid averagevelocity, U,l
the head pot and feed tank, were accounted for by adjusting the Ul;most air lift pumps operate at low liquid velocities, with slip
geometric submergence ratio to an effective submergence ratio. ratios between 1.5 and 2.5 (Stenning and Martin, 1968). The air
The equation found better agreement with data than previous voidage, E , may be found from the relationship
design approaches used in fuel reprocessing, but cannot be used in W,lW, = U,(1 - €)/U,E (3)
large bore air lift design outside of the nuclear fuel reprocessing
field, since correlations for A and B have been proposed only in where W, and W, are the air and liquid superficial velocities in
the 12 to 50 mm pipe diameter range. the pipe. In practice, the air void fraction is accurately predicted
Although many methods are available for predicting the liquid by using the drift-flux model described by Nicklin et al. (1962)
flow rate in air lift pumps, none of the equations is both theoreti- and by Zuber and Findlay (1965).
cally sound and accurate in prediction of pump performance. W& = C,(W, + W,) + V” (4)
What is required is a general analysis that can be applied with
confidenceto any air lift pump configuration. where C, is a constant accounting for the location of the slug in
the faster region of flow, and V, is the drift velocity between the
air phase and the total flow. The drift-flux model has been used
ANALYSIS extensively to predict holdup in a variety of vertical two phase
flows (Clark and Flemmer, 1984,1985; Ardron and Hall, 1980;
An air lift design equation is derived below, using a momentum Clark, 1984b; Miller et al., 1984). A range of values for C, in bub-
balance based on established two-phase flow theory. This equa- ble and slug flow have been reported in the literature to date
tion is derived making assumptions that are valid for the practical (Ishii, 1977; Ishii and Grolmes, 1978; Clark and Flemmer, 1984,
operating range of air lift pumps, and is not intended to be ap- 1985), and recently Miller et al. (1984) have presented values for
plied without consideration to all vertical two-phase flow situa- C, in gas-slurry slug flow. Although C , may not be predicted uni-
tions. versally, the best value currently available for slug flow is
Consider a vertical pipe partly submerged in a reservoir of wa- C, = 1.2. Data have also been presented by Shipley (1984) to
ter, and let the base of the tube have a reference height of zero show that a value of C, = 1.2 can be used for two-phase flow in
(Figure la). If the tube is submerged a depth of x1 below the large diameter pipes. Drift velocib V,, is given by the formula
surface of the supply vessel, then the static pressure, Po, at the
base of the pipe, is given by V, = 0.35(gd)0.5 (5)
where d is the pipe diameter.
Frictional loss in slug flow is less readily predicted. However,
where el is the liquid density Neglecting entrance effects, Po is because air lift pumps operate at relatively low liquid velocities,

Page 58 January, 1986 AlChE Journal (Vd. 32, No. 1)

the hydrostatic head is much larger than the frictional head loss air voidages below 50 % ,as encountered in the efficient operating
and an approximate relationship for the frictional contribution range of air lift pumps, the graphical Lockhart and Martinelli
will suffice. For example, in most air lift pumps frictional losses correlation is well represented by an expression of the form
represent only 2 to 10 % of the total head, so that the use of a sirn-
F = D(l + nt) (10)
ple pressure-loss expression is justified in the interests of obtaining
a closed-form design equation, and will cause little loss of accu-
racy. Models to predict pressure loss in slug flow are reviewed be-
-
In this range of air voidage, using n 1.5, Eq. 10 deviates from
the Lockhart and Martinelli correlation typically by 1 or 2 % and
low. Nicklin et al. (1962) have argued from continuity that the ve- at most by 6 % .Moreover, Eq. 10 does not deviate significantly
locity of the liquid slug in the pipe is equal to the total superficial from the equations proposed by Nicklin et al. (1962)and Stenning
velocity, (W, + W l ) .Since the liquid slugs are in contact with a and Martin (1968) over the effective range of air lift pumps. Ac-
fraction of approximately (1 - E ) of the pipe wall, this suggests cordingly, Eq. 10was chosen to describe the frictional losses in the
that the frictional loss per unit length due to the liquid slugs is pump. Although it is acknowledged that more sophkticated
given by models might be adopted to predict the slug flow pressure loss,
these will cause only a small increase in overall accuracy for prac-
F = 4eJ(W1 + WJ2(1 - ~ ) / 2 d (6) tical air lift pump designs, and will prohibit the development of a
where the friction factor is found for the total flow velocity using design equation in closed form. Combining Eqs. 2 and 10 results
the liquid phase properties. The same equation was reached by in:
Orkimski (1967), while Griffith and Wallis (1961) proposed - dP = [elg(l - E) + D(l + n ~ ) ] d x . (11)
that
Next, Eq. 4 for the prediction of air void fraction is substituted
F ...[4@1.f(wi+ W,)2/2dI[Wi/(Wi + WJl (7) into the total pressure differential, Eq. 11. Clearly, the superficial
air velocity in Eq. 4, W,, will vary with pressure over the height of
Stenning and Martin (1968) chose the Griffith and Wallis equa- the pump, and account must be taken of this variation in the total
tion to describe pressure loss in their analysisof air lift pumps, but differential. It is valid to mume that the air behaves in an ideal
found experimentally that the friction factor was larger than the fashion, and that expansion of gas in the pump is isothermal.
friction factor diagram predicted, although the flow in their short Where the mass flow rate of air to the base of the pump is G, the
length of pipe may not have been fully developed. A disadvantage superficial air velocity, W,, is given by
of these simple correlations when frictional losses must be known
precisely is that none,takes into account the shear stress generated W, = GPJAPe, (12)
at the wall by the annular film of liquid around the bubble. How- where Po is atmospheric pressure, A is the pipe cross-sectional
ever, in air lift pump design this contribution to the overall head area, and egisthe densityof the air at atmosphericpressure and at
may be considered negligible. the temperature of the pump. Setting M = GPJAp,, and com-
The approach of Lockhart and Martinelli (1949) has gained bining Eqs. 4, 11, and 12,
wide acceptance. Although not the most sophisticated approach,
it provides a simple expression sufficiently accurate for air lift
analysis. Irreversible two-phase pressure loss in the pipe is given
by the product of a two-phase flow multiplier, 4:, and the fric-
tional pressure loss per unit length of pipe, D , that would occur if
- dP = [ @Ig('

D(l +
- C,(M +
W , P ) + VJ
nM
C,(M + WIP) + VJ'
+
" >
the liquid alone were flowing in the pipe.
Multiplying through by the term C,(M + W,P) + V,P, and col-
F = 04,' (8) lecting pressure terms on the lefthand side of the equation,

cfl + (C,>W,+ V")P dP + dx = 0 (14)

(elgCoA4 - el@ +DC,,M + DnM + (e,g + D ) (COWl+ VJP

The pressure loss that would occur with the liquid flowing alone is
given by
Integrating between the points (x = 0, P
p = PZ),
= P,,) and (x - x2,

D = 4eJWIe/2d (9) whereR = elgM(Co-l) + D M ( C , + n )andS = C , W , + V , .

Since every variable except x, is known in Eq. 15, an explicit equa-
wherefis the friction factor, found from a conventional diagram. tion for the evaluation of x, has been developed, thus avoiding the
Values for the two-phase flow multiplier are obtained graphically use of incremental methods in the design of air lift pumps. The lift
from a correlation with a factor Xz,defined as the ratio of the pres- of the pump is then given by x, - xi, and the submergence ratio,
sure loss that would occur if the liquid were flowing alone in the or ratio of submergenceto total pump length, by x,Ixz.
pipe, to the loss that would occur if the air were flowing alone in Although Eq. 15 can be used directly to evaluate the lift of the
the pipe. Oshinowo and Charles (1975) found that frictional pump, given flow rates and pump submergence, it cannot be re-
losses in slug flow were slightly lower than are predicted by the arranged in closed form to give the liquid flow rate as a function
Lockhart and Martinelli correlation, and that at low liquid veloc- of gas flow rate if the lift is already known. Therefore, to generate
ities the direction of shear stress at the wall could even reverse due an operating curve Eq. 15 must be used on a trial and error basis,
to the liquid flowing back around the rising slug. However, when adjusting the liquid flow rate for a given gas flow rate until the
it occurs, this negative shear is very small relative to the hydro- predicted lift matches the actual pump lift. However, this method
static head, and deviation of the Lockhart-Martinelli prediction is still rapid in use, and when applied to tall pumps is superior to
in this region would not affect the overall air lift pressure gradient earlier momentum balances, which required an incremental
significantly. Of all the modelspresented above, that of Lockhart analysisover the length of the air lift tube as well as a trial and er-
and Martinelli was chosen to predict frictional pressure loss. For ror test.

AlChE Journal (Vol. 32, No. 1) January, 1986 Page 59

 Experimental operating curves of liquid flow rate vs. gas flow
rate for air lift pumps demonstrate that the same liquid flow rate
: 20
>t t 20%
can often be obtained using two different air rates. Liquid flow -
rate rises with increasing gas flow rate, through a point of maxi-
mum energy efficiency, to a point of maximum liquid flow rate 15-
(Figure 2). Thereafter, higher has superficial velocities in the
pump cause increasing frictional losses, and the liquid flow rate
falls off. In this region of high gas flow, exit effects, flow accelera-
10-
tion, and entrance effects at the gas sparger or orifice become sig-
nificant. Moreover, the pump can sometimes operate in annular
flow or very irregular pulsating flow rather than slug flow in this
region, so that holdup will not be predicted by the analysis used 0 5-
above. Not only is prediction of the pump performance difficult Data of Shaw - San
Fernando airlift
at very high gas rates, but no pumps are operated in this region in 0 127m Diameter
riveted steel
practice. Most pumps are operated at or near the point of maxi-
mum efficiency on the rising section of the operating curve.
0 05 10 15 20
Equation 15is therefore intended to predict the first (low gas rate)
solution only, since this is the only solution of interest in industrial MEASURED WATER SUPERFICIALVELOCITY WI (m/sec.)
air lift operation.
Figure 3. Comparison of the tali air lift data of Shaw (1920)
with the prediction of Eq. 15.
COMPARISON OF ANALYSIS WITH EXISTING DATA
steel, which has a high equivalent roughness, so that a friction
The most stringent test for the theory developed above is factorof0.01 waschosenastypicalfortheflow. Avalueof 1.2was
whether or not it can predict the behavior of very tall air lifts, used for the profile constant, C,, the exponent n was taken as 1.5,
where the gas volumetric flowrate changes significantly over the and the drift velocity, V,,,was calculated at 0.39 mis. Agreement
height of the pump. For short air lifts, it is sufficient to use an av- between Shaw’s data and the new design equation is good, with
erage gas flowrate in the pump (Stenningand Martin, 1968), but all predictions but one falling within the 20 % lines on the parity
this approach is not valid for taller pumps. Shaw (1920) has sup- plot, Figure 3. However, it is suspected that the one point in dis-
plied data on a variety of tall air lifts (up to 400 m high) used to agreement may be reported in error, since it does not follow the
dewater the Tiro and San Fernando mine shafts in Mexico. Some trends in the rest of Shaw’s data. Small errors in predicting the re-
of Shaw’spumps had variable diameters, so that they could not be maining points may be due to a phase change from slug to annu-
compared with Eq. 15, but the pump used in the San Fernando lar flow near the top of the San Fernando pumps, since air super-
shaft was of constant 5 in. (127mm) diameter so that it was suited ficial velocities were very high in this region as a result of the low
to comparison with the theory. (Variable diameter can offer in- pump submergence used. The theory developed abwe was for
creased efficiency for very high single-stage lifts, but is seldom slug flow, and will become inaccurate in predicting annular flow
used in modern designs.) In Table 13 (p.436) of Shaw’s paper, op- behavior, in particular the liquid holdup in annular flow.
erating data for five different submergence ratios and heights of Comparison with Shaw’s data has demonstrated that the new
the San Fernando lift are supplied. For each configuration, data design equation is effective in predicting the performance of tall
are given for the point of maximum flow rate for the pump, and air lift pumps, and that a value of C,, = 1.2 is satisfactory for use
for a point in the range of higher operating efficiency, so that a to- in large diameter air lifts. Experimental data described below
tal of ten test points was available. were used to test its accuracy in the design of mid-size air lift
Shaw’sdata compare favorably with the predictions of the new pumps used in nuclear fuel reprocessing.
equation; as shown in Figure 3. The pipes were of spiral riveted
EXPERIMENTAL

Tests were undertaken using an air lift installation with a 38.1 mm dia.
tube, describd in detail by Dabolt and Plummer (1980), and diagamed
in Figure 4. The lift was constructed of schedule 40 stainlesssteel pipe, and
gas inleb were situated so that submergence ratios of 60,70, and 80%
could be chosen. The head pot of the pump was constructed of plexiglas
and stainless steel, and contained two baffles and a weir for measuring
flowrate of liquid. Air supply was measured using a rotameter.
MAXIMUM
EFFICIENCY A276L tankwasusedas thefeedreservoir, andtheliftdischargedliquid
via the head pot into a smaller 62 L reservoir. Liquid was lifted back to the
larger reservoir by a separate 50 mm dia. air lift pump. In addition, the
equipment vent lines wereconnected via PVC piping to the off-gasheader,

I
P with valves and pressure gauges installed in each line. This allowed either
3
pressure or vacuum to be applied to each piece of equipment as desired.
2 Transition to Annular Eleven test runs, of approximatelyten points each, were made to deter-
Flow can occur mine the effects of submergence, liquid properties, and vacuumlpressure
on the system. Data were gathered for the ascending section of the air lift
operating curve, since operating points to the right of the point of maxi-
mum flow are of no practical interest to the air lift operator. Conditions for
GAS VOLUMETRIC FLOWRATE the runs are listed in Table 1. Three nonaqueous liquids used in the runs
were 30 70 TBP in dodecane, the same solution saturated with sodium car-
Figure 2. Typical air lift pump operating curve. Equation 15 Is bonate, and with nitric acid.
valid for the ascending portion of the curve, up to the Air flow rate was measured accuratelywith a rotameter, while expected
maximum liquid flow rate. errors in liquid flow rate measurement are shown by error bars in Figure 5.

Page 60 January, 1986 AlChE Journal (Vol. 32,No. 1)

 I .o
Submergence Ratto In parentheror -

r-
15- -J 0.8-
x
-E
\

cG 0.6-
-
ea 'W3
>
-
I-
10
2
-I

S ii 0.4-
2
w
a
UJ
I n
3
v)
9
=
P
0.2-
5 -I

I
I . 1 I I
0 0.2 0.4 0.6 0.8 10

FREE AIR SUPERFICIAL VELOCITY (rnlsec.)

0
Figure 5. Comparison fo test data for water runs wlth
Flgure 4. Experlmental test instaiiatlon. operating curves generated using the new design equatlon.
A. 276 Ltank 1. 19 mm air supply
8. Metering head pot 2. 38 mm liquid supply
C. Separator 3. 50 mm air lift
the operating range was found to produce no significant change
D. 62 L tank 4. 19 mm air supply in the CUN~S, because the frictional losses were small at the low
E. Glass viewing sections 5. 38 mm experimentalair lift liquid velocities used. This amplifies the claim that very accurate
F. Vents 6,7. 38 mm drain from head pot frictional loss models are not required for practical air lift design.
G. Drains 8. 150 mm horizontalfeed to tank
H. Pointsof air introduction 9. 50 mm tank drain
Also, losses due to the liquid flow in the downcomer were small,
and were neglected.
Agreement between the data and model is good, except in the
case of run 2-8, where trends in the data suggest an operating
RESULTS curve with a lower slope than is predicted. However, noting the
operating conditions, it is clear that the curves for runs 2-1 and 2-8
Points on operating curves for the runs, as plots of iiguid flow should not cross at any point, while the data suggest that they do.
rate vs. free air flow rate, are shown in Figures 5 to 9. In each case It was concluded that the first three experimentalpoints for run 2-
the given curve is predicted using the new design equation, with a
friction factor of 0.01, n = 1.5, C, = 1.2, and the drift velocity
V, = 0.21. It should be stressed that each of these constants is
known prior to obtaining the-experimental results, and is not
found by regression. The friction factor of 0.01 was chosen as rep-
resentative ofO the flows from a conventional friction factor dia- "
gram. More accurate evaluation of the friction factor throughout -
4 0.8-
\
Y)

-E
TABLE1. FOR EXPERIMENTAL
CONDITIONS RUNS c
V
9 0.6-
Run Submer- W
>
No. gence, % Pressure Fluid A
~~
9
2- 1 82 Atmospheric Water u
0.4-
2-2 82 - 25.4 cm water Water W
2-3 70 - 25.4 cm water Water n
2
2-4 60 - 25.4 cm water Water v)

Q Submergence Rat10in parenthesis

2-5-1 82 - 25.4 cm water Solvent*
3
25-2
2-6-1
82
82
Atmospheric
- 25.4 cm water
Solvent'
Solvent + carbonate**
+ carbonate**
1
0.2-
.
RUN 2-5-1(82%)
R U N 2-5-2
(82%)
2-6-2
2-7-1
2-7-2
82
82
82
Atmospheric
- 25.4 cm water
Atmospheric
Solvent
Solvent + nitric acidt
Solvent + nitric acidt
0 L----- 0.2 0.4 0.6 0.8 3

FREE AIR SUPERFICIAL VELOCITY (m/sec.)

'30% TBP in dodecane, sg. 0.8227.
"Solvent saturated with carbonate, sg. 0.8295.
tSolvent saturated with nitric acid, sg 0.8305.
Figure 6. Comparison of test data for solvent-only runs wlth
operating curves generated using the new design equation.

AlChE Journal (Vol. 32, No. 1) January, 1986 Page 61

 1 .o
1
’ O - ’
I

v
\
VI 0.8-
-E >
6
cu 0.6 -
9W
> 9
-
u
U
a
0.4- W
a
2
v)

P
Submergence Ratio m parenthesis 3
P a
3i 0.2- RUN 2-6-1 (82%) J

RUN 2-6-2 (82%)

I I I I 1
0 0.2 0.4 0.6 0.8 1.o

FREE AIR SUPERFICIALVELOCITY (m/sec.) FREE AIR SUPERFICIAL VELOCITY (m/sec.)

Figure 7. Comparison of test data for solvent carbonate + Flgure 9. Comparison of test data with Operating curves
runs with operating curves generated using the new generated using the new design equation.
design equation.

8 represent too high aliquid flow rate, although the cause has not COMPARISON OF DATA WITH OTHER MODELS
been identified.
In many of the experimental runs, the liquid flow rate was Several other design equations were also compared with run 2-
slightlylower than predicted over the mid-range of the data. This 1, as shown in Figure 10. The Husain and Spedding (1976) analy-
depression in the operating curve has not been fully explained, sis predicted too low a liquid flow rate when solved for the operat-
but might be attributed to the stability of the two-phase flow in ing conditions in run 2-1, using the constantr supplied by Husain
the U-tube apparatus. Theory of pump stability has been dis- and Spedding in their paper. It was deduced that although the
cussed mathematically by Hjalmars (1973) and Apazadis (1980,
1982, 1983), but has not been related directly to such perturba-
tions in the operating curve. However, the overall agreement be-
- THIS DESIGN EQUATION
............. HUSAIN & SPEDDING
tween theory and data was good, thus demonstrating that it is - -- DABOLT & PLUMMER EMPIRICAL EON.
possible to predict pump performance without resorting to re-
gression on existing pump data.
-- MODIFIED ENERGY BALANCE (O’NEILL)
.............
ENERGY BAIANCE (100%EFFICIENCY)
0 DATA RUN 2.1
10.

0 8-

0.6-

0.4-

Submergence Ratio on parentheas

0 2- RUN 2-7-1 (82%)

RUN 2-7-2 (82%)

0 012 0‘4 016 018 1 lo

0 0.2 0.4 0.6 0.8 1.o FREE AIR SUPERFICIAL VELOCITY (m / sec)
FREE AIR SUPERFICIAL VELOCITY (m/sec.) Figure 10. Comparison of several design equations with the
Figure 8. Cornparison of test data for solvent nitric acid + data of run 2-1.
The ideal energy balance (100% efficiency) and Husaln and Spedding
runs with operating curves generated using the new method are both inaccurate. The Ingersoll-Rand equation presented by
design equation. O’Neill(l975) cannot predict the operating curve except as a straight line.

Page 62 January, 1986 AlChE Joumal (Vol. 32, No. 1)

form of the Husain and Spedding equation might reflect the air D - frictional pressure loss per unit length in single phase flow,
lift performance, the terms in the equation were not necessarily Palm
constant over a range of pump geometries. Moreover, the Husain d = pipe diameter, m
and Speddinganalysis is in error at very low gas flow rates, where F - frictional pressure loss per unit pipe length in two-phase
it still predicts a very small liquid flow rate; in other words, the flow, Palm
predicted operating curve does not cross the horizontal axis at ri f - friction factor
positive value for air flow rate. A correct solution will predict an G - m a s flowrate of air, kg/s
operating curve which does cross the horizontal axis into the g - acceleration due to gravity, m/s2
fourth quadrant, this region corresponding to a hypothetical M = a product of gas superficial velocity and pressure, Pa*m/s
downflow of liquid in the pump, if the whole length of the air lift n = constant
tube is to remain full of a gas-liquid mixture. Only the new equa- P - pressure, Pa
tion developed above predicts this situation correctly. R - relationship extracted for clarity in Eq. 15
The Ingersoll-Rand equation (ONeill, 1975) agrees quite fa- s = relationshipextracted for clMty in Eq. 15
vorably with the data, particularly in the region of maximum ef- u = average velocity Over pipe cross section, mls
ficiency However, since it is a modified energy balance with a V“ = drift velocity of slug, m/s
constant efficiency assumed from operating experience, the oper- W - average superficial velocity over pipe cross section, m/s
ating curve is a straight line which becomes inaccurate at very x - height above base of pump, m
high or very low gas flow rates. € = gas void fraction
The empirical equation of Dabolt and Plummer (1980) and the
new equation developed above find superior agreement with the
experimental data in Figure 10.
4; -
e - density, kg/m3
two-phase flow multiplier

Subscripts
DISCUSSION a - atmospheric
The new model agrees with experimentalresults and data from
the literature. Equation 15 reduces correctly to the single-phase
g
1
o
--
= air,gas
liquid
at gas sparger (except C,)
pressure differential when the air rate is zero, and has found fa- w = water
vorable comparison with the empirical Ingersoll-Rand equation
(O’Neill, 1975) in the region of maximum efficiency. However,
the Ingersoll-Rand equation predicts only the volume of air re-
1
2
=
- at liquid level in reservoir
attopofpump
quired to raise a given volume of water, given the lift and submer- LlTERATURE CITED
gence, and so cannot take into account frictional losses, which
may become significant in pipes of smaller bore. In addition, the Apazadis, N., “The Stabilizingof an Airlift Pump by Flow Reduction:’ Re-
slip velocity will differ between the case of a high flow-velocity port, Dept. Mechanics, Royal Inst. Tech., Stockholm (1980).
pump with a small pipe bore, and a low-velocity large bore .
- “Influenceof Bubble Expansion and Slip on the Performanceand
pump, although the volume rate of liquid lifted may be identical Stability of an Airlift Pump,” Report, Dept. Mechanics, Royal Inst.
in each case. The Ingersoll-Rand approach cannot account for Tech., Stockholm (1982).
thisdifference. Equation 15takes into account both the variation .
- “Influenceof Friction on the Efficiency and Stabilityof an Airlift
J?ump,”Report, Dept. Mechanics, Royal Inst. Tech., Stockholm (1983).
of slip velocity and the presence of frictional loss, and may there- Ardrun, K.H., and P.C.Hall, “Predictionof Void Fraction in Low Velocity
fore be applied over a wide range of operating conditions for air Vertical Steam-WaterFlow,” ASME I Heat Tmns., 102,3 (1980).
lift pumps in slug flow. Brown, K.E., Gas Lij? Theory and Practice, Petroleum Pub. Co., Tulsa
The linear assumption for the relationship between pressure OK 23-306 (1967).
loss and voidage, Eq. 10, will cause negligible error in the final Clark, N.N., Air Lijt Pumpsfor Hydmdic Tmnsportof Solids, Powder Ad-
prediction of lift. Equation 10 deviates from the widely used visory Centre, London, 13-17 (1984a).
Lockhart and Martinelli (1949) model by no more than 6 % at air -. “Predicting the Circulation Rate in Pachuca Tanks with Full
voidages up to 0.5. Moreover, the frictional loss in typical air lift Height Draft Tubes,” Min. Metal. Proc. (Trans. A I M E ) 276, 226
pumps is at least an order of magnitude smaller than the hydro- (1984b).
Clark, N.N., and R.L.C. Flemmer, “On Vertical Downward Two-Pha$e
static head, so that overall errors incurred by this linear assump- Flow,” Chem. Eng. Sci., 39,170 (1984).
tion would total no more than 0.5 % . In pipes of smaller bore, for - , “PredictingHoldup in Two-PhaseBubble Upflow and Downflow
examplethe extreme case of a 50 mm pipe carrying a flow at 2 to 3 Using the Zuber and Findlay Drift-Flux model,” AlChE J . , 31, 500
mls, the frictional loss is only of the same order of magnitude as (1985).
the hydrostatic head, so that the error incurred by the linear as- Dabolt, R.J., and N.N. Clark, “Pumping Radioactive Slurries by Air Lift,”
sumption would be less than 3 % . 10th Powder and Bulk Solids Cod., Rosemont, IL, International Pow-
Using the new equation, air lift pumps are rapidly designed. In der Inst., London, 779 (1985).
conjunction with an equation describing the efficiency of the Dabolt, R.J., and K.E. Plummer, “Design of Air Lift Systemsfor Transfer
pump, such as that presented by Nicklin (1962), the closed soh- and Measurement of Radioactive Liquids,” Dept. of Energy Report
AGNS-35900-3.2-77(1980).
tion developed in this article provides an efficient means for opti- Govier, G. W., and K. Aziz, The Flow of Complex Mixtures in Pipes, Van
mizing such variables as pipe diameter, gas flow rate, and sub- Nostrand Reinhold, New York, 389414 (1972).
mergence ratio. The design of a pump with minimum energy G~iffith,P., and G.B.Wallis, ‘‘Two-Phas Slug Flow,” ASME J . Heat
requirements by using this method is more rapid than by using in- Trans., 83,307 (1961).
cremental methods, and more accurate than by using existingem- Hjalmars, S., “The Origin of Instability in Airlift Pumps,” ASME J Appl.
pirical equations. Mech., 95,399 (1973).
Husain, L.A., “On the Gas-Lift pump: A New Approach,” Paper G2,Znd.
Symp. Jet Pumps, Ejectors, and Gas Lift-Techniques, Cambridge
NOTATION (Mar., 1975).
Husain, L.A. andP.L. Spedding,“TheTheoryoftheGasLiftPump:’Int.
A = pipe cross-sectional area, m2 J. Multiphase Flow, 3,83 (1976).
C, = constant for the drift-flux model Ishii, M., “One-Dimensional Drift Model and ConstitutiveEquations for

AlChE Journal (Vol. 32,No. I ) January, 1986 Page 63

 Relative Motion Between Phases in Various Two-Phase Flow Regimes,” O’Neill, P.P., “Pumping of Liquids and Gases,” Chemical Engineers Hand-
ArgonneNational Lab. Rept. 77-47 (Oct., 1977). book, Perry and Chilton, Eds., McCraw-Hill, New York,6-15, (1975).
Ishii, M., and Gmlmes, “Constitutive Equation for One-Dimensional Oshinowo, T., and M.E. Charles, “Vertical Two Phase Flow.’’Can. J.
Drift Velocity of Dispersed M P h a s e Flow,’’Two-Phase lfansport and Chem. Eng., 52, 438 (1979.
Reactor Safety, T.N. Vezimglu and S. Kakac, Eds., Hemisphere, Wash- Orkizewski, J., “Predicting Two-Phase Pressure Drops in Vertical Pipe,” J.
ington, DC (1978). Petrol. Tech.,19,829 (1967).
Jeelani, S.A.K., K.V. Kasipati Rao, and G.R. Balasubramanian, “The Ros, N.J.C., “SimultaneousFlow of Gas and Liquid as Encountered in
Theory of the Gas Lift Pump: A Rqoinder,” Int. J. Mulliphase Flow, 5, Well ntbing,”J. PetruZ. Tech., 13,1,037 (1961).
225 (1979). Shaw, S.F., “UnwateringtheTiro General Mine by.4ir Lift,” Tram AIME,
Lawson, J.D., and J.P. Brill, “A Statistical Evaluation of Methods Used to 421 (1920).
PredictPressure Laws for MultiphaseFlow in Vertical Oilwell lbbing,” Shipley, D.G., “Two-Phase Flow in Large Diameter Pip:’ Chem. Eng.
ArEificiol Lift SPE Rep. Sex, 84-95 (1975). Sci., 39, 163 0984).
Lockhart, R.W., and R.C. Martinelli, “Proposed Correlation of Data for Stenning, A.H.,‘and C.B. Martin, “AnAnalyticalandExperimentalStudy
Isothermal Two-Phase, Two-Component Flow in pipes,” Chem. Eng. of Air Lift Pump Performance,” ASME J . Eng. for Power, 90, 106,
Pmg., 45,39 (1949). (1gw.
Miller, R.L., M.B. Cain, and M.P. Sharma, “Flow Regime Transitions in Taitel, Y., D. Barnea, and A.E. Duckler, “ModelingFlow Pattern Transi-
Vertical Upward Threephase Gas-Liquid-Solid Flow,” Paper l a g , tions for SteadyUpward Gas-Liquid Flow in Vescal Tubes;’ AIChE J.,
AIChE Ann. Meet., San Francisco, (1984). 22,47 (1980).
Nicklin, D.J., “The Air Lift Pump: Theory and Optimization,” Tmns. Zuber, N., and J.A. Findlay, ‘Xverage Volumetric Concentration in Two-
IChE, 41,29 (1963). PhaseFlow Systems,”ASMEJ. Heat Illam., 87,453, (1965).
Nicklin, D.J., J.O.Wilkes, and J.F.Davidson, “Two-PhaseFlow in Vertical
Tubes,” Tins. IChE., 40,61, (1962). Manuscript received Oct. 3, 1984, and revision received Mar. 14, 1985.

Page 64 January, 1986 AlChE Journal (Vol. 32, No. 1)