Venturi Scrubber Optimization Revisited
Venturi Scrubber Optimization Revisited
Venturi Scrubber Optimization Revisited
To cite this article: Douglas W. Cooper & David Leith (1984) Venturi Scrubber Optimization
Revisited, Aerosol Science and Technology, 3:1, 63-70, DOI: 10.1080/02786828408958994
A previous paper of ours on venturi scrubber optimiza- suggest that the scrubber geometry affects the effective-
tion used a model by Calvert et al. to describe particle ness with which liquid is utilized and the effectiveness
collection. Several assumptions in this model are im- with which energy is utilized. Furthermore, the conditions
proved in the present work; actual values for droplet drag necessary to obtain optimum venturi performance change
coefficient are used rather than the Ingebo approxima- with particle diameter, throat velocity, and scrubber
tion, and the actual scrubber geometry is used rather geometry. With these results, the conditions necessary
than assuming the scrubber is a duct infinitely long and for theoretically optimum performance of the scrubber
constant in diameter. Results from the present model can be readily identified.
METHODS
We prepared a mathematical model to en- LABORATORY SCRUBBER
able us to investigate theoretically the effects
of scrubber geometry on scrubber perfor-
mance. This led to a study of the choices
others have made for droplet drag coefficient "GRADUAL" SCRUBBER
and to a general review of the theory un-
derlylng such models. The model we de-
veloped is similar to that of Boll (1973). It J
1
assumes one-dimensional gas flow in the axial
"ABRUPT" SCRUBBER
direction, with velocity equal to the gas volu-
metric flow rate divided by the duct cross-
sectional area:
We used the expression for particle collec- the gradual geometry here, had the same
tion by impaction used in the model of throat dimensions but a diverging section
Calvert et al. (1972), 1.34 m long, as long as the diverging and
final duct sections of the laboratory scrubber
combined. The third scrubber geometry,
in whlch qd is the ratio of the cross-sectional called the abrupt geometry here, had a throat
area of the gas swept clean by a droplet to the same dimensions as the other two
the cross-sectional area of the droplet, and K scrubbers, but no diverging section; the
is the impaction parameter: throat emptied directly into the 0.12-m-diam
duct. The total length for each of these
scrubbers was 1.40 m. Additional modeling
The particle relaxation time r is (Fuchs, was conducted for a scrubber of effectively
1964): infinite length, modeled as a straight pipe
with diameter 0.056 m and length 50 m.
Drag Coefficient
Scrubber Geometry
We evaluated several relationships for the
Figure 1 shows the scrubber geometries used drag coefficient with this model. The Ingebo
in this study. First is the geometry of the (1956) relationship used in the model of
scrubber we are using for laboratory experi- Calvert et al. (1972) is
ments now in progress. This laboratory
scrubber had a throat 0.056 m in diameter
and 0.06 m long attached to a diverging
section 0.44 m long, which connected the A different relationship for C, was pre-
throat to a duct 0.12 m in diameter and 0.90 sented by Hollands and Goel (1975) and
m long. The second scrubber geometry, called used by Yung et al. (1977) in a revision of
Venturi Scrubber Optimization Revisted
the model of Calvert et al. (1972): FIGURE 2. Predicted gas velocity, droplet veloc-
ity, and volume swept clean per volume of liquid
c, = c,,,/&, (7) versus distance along scrubber, measured from
in whch C,,, was evaluated at the throat scrubber throat inlet.
conditions from the "standard" curve given
by Dickenson and Marshall (1968): coefficient. With this baseline information,
we could determine the extent that predic-
C, = 0.22 + (24/Re)(l + 0.15Reo6). (8)
tion of the scrubber performance is affected
The rationale for extrapolation from Eq. (8) by assumptions about the scrubber geometry
seems questionable, however, when (8) itself and drag coefficient. The present model con-
can be used. Boll (1973) also developed a firmed that the 50-m scrubber would give
model for venturi performance based on the best performance at the conditions previ-
standard curve to determine the drag coeffi- ously predicted (Leith and Cooper, 1980),
cient. Most of our work also used the stan- fK = 1, where f is an empirical factor some-
dard curve, as discussed later. what smaller than 1 (often f = 0.5).
In contrast to the 50-m scrubber, the
laboratory scrubber geometry had best per-
RESULTS
formance for fK near 10, for a throat velocity
Comparison with Previous Theoretical Results of 61 m sec-'. Even larger values of fK were
optimal for our laboratory scrubber geome-
Our model, using (6) for the drag coefficient, try at velocities higher than 61 msec-l. The
was used to compare optimization predic- difference in these predictions shows the
tions for the scrubber with 50-m throat weakness of models that assume infinite
against the predictions by our model for the throat length.
laboratory scrubber geometry. Our previous
theoretical work (Leith and Cooper, 1980)
was based on the model by Calvert et al. Droplet Motion Through Scrubber
(1972), which uses (6) for the drag coeffi- The remaining work reported here used (8)
cient. We wanted to determine whether our for the drag coefficient. Figure 2 shows typi-
new model would behave the same as the cal results. Here, the droplet diameter was 50
Calvert et al. (1972) model under the same ym; the particle aerodynamic diameter was 1
assumptions regarding the geometry and drag ym. The horizontal axis in this figure is the
66 Douglas W. Cooper and David Leith
distance from the throat entrance. The left penetration Pn of particles of a given size is
vertical axis is the droplet or gas velocity.
Pn = exp(- v*/vG). (10)
The right vertical axis gives the ratio of gas
volume each droplet sweeps clean of par-
ticles to the volume of each droplet, V*/VL, Here the argument of the exponential func-
called the "specific cleaning volume" by tion is the ratio of the volume swept clean to
Loffler and Schuch (1981). The volume swept the volume of gas:
clean, V*, is the integrated product of the
droplet collection efficiency q,, the droplet
cross-sectional area A,, and the distance the The second set of parentheses contains the
droplet travels through the gas: liquid-to-gas ratio, so that the values of the
specific cleaning volume, V*/VL, from Fig-
V * = I t q , ~ , l u d - uGjdt. ures 2-4 can be used to estimate the penetra-
0
tion once a liquid-to-gas ratio is selected.
Figure 2 shows that the gas velocity in
the throat was 61 msec-' and slowed to
13 msec-I in the final straight section of the Geometry, Particle Size, Gas Velocity, and
scrubber. The droplet velocity increased Liquid Utilization
rapidly from zero in the initial portion of the Results from simulations like that in Figure
scrubber, driven by the high relative velocity 2 are compiled in Figure 3 for different
between gas and droplet. At a point about scrubber geometries, gas velocities, and
0.2 m downstream of the throat entrance in droplet diameters. The particle diameter was
this case, the gas and droplet velocities were fixed at 1 pm. Figure 3 displays the ratio of
equal. From there on, the droplet velocity the volume of gas swept clean by a given
decreased because, from there on, the gas droplet per unit droplet volume as a function
velocity was less than the droplet velocity. of droplet size for three different throat
The gas volume swept clean of particles, velocities and for several scrubber geome-
V*, increased with the droplet efficiency, tries.
which in turn increased with the relative Figure 3 shows that the specific cleaning
velocity between droplets and gas. In the volume V*/VL increased strongly with gas
venturi throat, the relative velocity between throat velocity. At 345 msec-' the throat
droplets and gas was especially high, and the velocity is sonic, requiring a pressure up-
gas volume swept clean, V*, increased stream twice that downstream. This situation
rapidly. At the point where the droplet and is rarely encountered but may have applica-
gas velocities were equal, no sweeping took tion to mitigating the effects of an accident
place and the plot of specific cleaning volume, at a nuclear power plant (Cooper, 1982).
V*/VL, is constant with distance. Addi- Analyses like that in Figure 2 determined
tional collection took place on the down- that particle collection occurred on the up-
stream side of the drop from the equal stream side of the drops as they accelerated
velocity point to the end of the scrubber, and in the venturi throat, and also on the down-
the volume swept clean increased accord- stream side of the drops as they decelerated
ingly. All together, this droplet swept a through the venturi diverging section and
volume about 1200 times its own volume. final duct. For each geometry considered
One-third of the sweeping occurred after the here, the venturi throat was identical. Be-
venturi throat. cause most drop acceleration occurred in the
If perfect mixing across the flow direction throat, collection during acceleration was
causes particle concentration to be uniform about the same for each geometry under
perpendicular to the direction of flow, the identical operating conditions.
Venturi Scrubber Optimization Revisted 67
-
between droplets and gas to be so high that
K > 10 and q, 1 for most of the droplet
deceleration. Although collection on the
the throat entrance, K, for particles 0.3, 1.0,
and 3.0 pm in diameter, all having the den-
sity of water, for the sonic velocity (345
downstream side of the droplets was effec- m sec-I), a typical velocity (61 m sec-l), and
tive, it occurred only immediately down- an unusually low velocity (10 m sec-l). These
stream of the throat, because the droplets calculations were done for the laboratory
decelerated quickly. In contrast, the gradual scrubber geometry. Other geometries would
scrubber geometry allowed gradual droplet give somewhat different results, as illustrated
deceleration over the entire length of the in Figure 3. Figure 4 shows some clear trends,
scrubber diverging section. At h g h throat but no simple relationship is evident for the
velocity, the relative velocity obtained was dependence of specific cleaning volume,
Douglas W. Cooper and David Leith
l0,OOO -
-
I I I
/p-: -
IMPACTION PARAMETER. K
V * / V L ,on impaction parameter and throat cur because, in the present model, limited
velocity. Unlike our previous model (Leith opportunity existed for sweeping. The actual
and Cooper, 1980), which assumed that Eq. gas volume swept clean was less than that
(6) describes the drag coefficient and that the possible in an infinite throat. This limited
venturi is adequately characterized as a throat opportunity was most important for larger
of infinite length, the present, more realistic, particles and higher velocities, which can
model showed that no single value for the better use a longer scrubber. Because the
impaction parameter always provides opti- extent of sweeping was constrained by the
mum collection. Note, however, that the val- scrubber geometry in the present model,
ues for the highest velocity, 345 msec-', the sweep efficiency became more important.
have been calculated with a theory that did As a consequence, high values of the impac-
not include the effects of gas compressibility, tion parameter were necessav to ensure
even though these effects may be important maximum target efficiency for each drop,
(Cooper, 1982). and curves of V * / V L against K were dis-
Our previous work (Leith and Cooper, placed toward higher K. Again, this displace-
1980) predicted that the specific cleaning ment was greater for larger particles and
volume V */ VLwould increase with increas- higher velocities, for which the scrubber
ing particle diameter d , and gas velocity u,, geometry constraints were more severe. The
although to much higher values than found conditions that led to most effective collec-
here. Furthermore, the previous model pre- tion cannot, apparently, be stated concisely.
dicted that maxima in plots of V * / V Lagainst The differencesbetween the "infinite" and
K occur at a constant value of K,, = l.l/f, the finite throat lengths were least for a
where f is the constant from the venturi collection of small particles at low gas veloc-
model of Calvert et al. (1972). The present ity. Under these conditions most collection
model predicted that the smallest value occurred in the venturi throat, as discussed
for K,, should be about unity for the small- in conjunction with Figure 3. In this case the
est values of d , and u,. At larger values scrubber geometry constraints were less im-
of d , and u,, K,, was displaced to higher portant; collection was reasonably modeled
values. by an infinite throat, as assumed by Calvert
Both a reduction in specific cleaning et al. (1972), and previous results based on
volume V * / V L and a displacement of the the Calvert model compare well with those
maximum V * / V L to higher values of K oc- given here.
Venturi Scrubber Optimization Revisted 69
Calvert, S., Lundgren, D., and Mehta, D. (1972). Ven- solid spheres in clouds accelerating in airstreams,
tun scrubber performance, J. Air Pollut. Control NACA Tech. Note 3762.
Assoc. 22:529-532. Leith, D., and Cooper, D. W. (1980). Venturi scrubber
Cooper, D. W. (1982). Venturi scrubbing for filtered optimization, Atmos. Enuiron. 14:657-663.
vented containment, 17th DOE Nuclear Air Clean- Loffler, F., and Schuch, G. (1981). Wet scrubber design,
ing Conf., Denver, CO, August. Fzltr. Sep. 18:70-74.
Dickenson, D. R., and Marshall, W. R. (1968). The rates Nukiyama, S. and Tanasawa, Y. (1938). Experiments on
of evaporation of sprays, AIChE J. 14:541-552. the atomization of liquids in an air stream. Trans.
Fuchs, N. A. (1964). The Mechanics of Aerosols. SOC.Mech. Engrs. (Japan) 436.
Pergamon, New York. Yung, S.-C., Calvert, S., Barbarika, H. F., and Sparks,
Hollands, K. G. T., and Goel, K. C. (1975). A general L. E. (1977). Venturi scrubber performance model,
method for predicting pressure loss in venturi Enuiron. Sci. Technol. 12:456-459.
scrubbers, Ind. Eng. Chem. Fundamentals 14:16-22.
Ingebo, R. D. (1956). Drag coefficients for droplets and Received 25 July 1983; accepted 20 November 1983