Single-Stage AC Electrostatic Precipitation

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 24, NO.

4, JULYIAUGUST 1988 717

Single-Stage AC Electrostatic Precipitation

Abstract-Measurements of the collection efficiency of a laboratory- required operating the precipitator within a narrow frequency
scale single-stage electrostatic precipitator (3-cm plate spacing) excited by range, high enough to avoid accumulation of sufficient charge
ac voltages (V,,, - 8 kV, f = 20 -+ 500 Hz) biased to give no time- to cause back-corona yet low enough to allow the particles to
average current are reported. With f = 60 Hz, an effective migration
velocity of 3.5 cm/s is observed. Previously published works on the
acquire significant charge levels.
charging and migration of particles in ac coronas, and on the effects of In Section 11, laboratory-scale experiments are described in
finite turbulent diffusivity on the precipitation process are combined to which the efficiency of an ac precipitator is measured as a
predict the experimentally observed dependence of the efficiency on the function of frequency and applied voltage. In Section 111 a
magnitude and frequency of the applied voltage. When applied to the model is proposed which predicts the collection efficiency in
experimental situation reported here, the model reduces to one of
complete mixing, with the effective particle migration velocity being
terms of the steady-state and oscillatory particle migration
predicted in terms of the velocity that is achieved with dc charging and a velocities, the effective diffusivity of the bulk flow, the
charging time constant that is either estimated or obtained from charging thickness of the laminar sublayer (within which no turbulent
experiments. diffusion takes place), and other parameters such as the plate
length, plate separation, and bulk flow velocity. The steady-
I. INTRODUCTION state and oscillatory particle migration velocities used can
either be taken as given parameters of the problem or
ITH the objective a class of electrostatic precipitators
computed using the ac corona-charging model presented by
that are relatively insensitive to the resistivity of the ash,
Ehrlich and Melcher [5].
measurements of the collection efficiency of a laboratory-scale
The model is simplified in Section IV, using the approxima-
single-stage ac electrostatic precipitator have been performed.
tions of large turbulent diffusivity (fully mixed flow) and a
These are described here and compared to theoretical predic-
laminar sublayer thick enough that the particle’s oscillatory
tions of the efficiency as a function of the magnitude and
migrations do not contribute significantly to their precipitation
frequency of the applied voltage. So that the dominant physical
rate. This limit is shown to result in a Deutsch model for the ac
processes can be identified, the model developed includes
precipitation process, with the effective particle migration
effects of finite turbulent diffusion.
velocity, weff, being replaced by the time-average migration
Lau [ 11 investigated the collection of highly resistive
velocity in the ac corona, W , , , ~ . Conclusions are drawn in
particulate [p - 1013 Q - cm] using ac voltages with
Section V.
frequencies in the range of 50-500 Hz. He suggested that the
use of bipolar excitations would allow an insulating cover to be 11. MEASUREMENTS OF EFFICIENCY
placed over the collecting plates so that sparks from wire to A. Description of Experiment
plates would be extinguished by the accompanying accumula-
Side and top views of the laboratory-scale precipitator are
tion of surface charge. Krug [2] recognized the possibility that
shown by Fig. 1. The corona wires extend 154 cm in the flow
the particles could reverse their charge each time the corona
changed its polarity, thus making nearly continuous progress direction, suspended midway between grounded metal plates,
which have a spacing of 3 cm. These form the top and bottom
toward the collection plate, with only momentary periodic
reversals of motion. Ehrlich [3] proposed that an ac precipita- walls of the duct. Glass sidewalls, which have a spacing of 14
cm, allow optical observation of the spatial distribution of the
tor could be run under conditions that would restrict the
accumulation of surface charge on the collected particulate corona. The corona wires are attached to a 0.64 cm brass rod
at the downstream end by springs, passed through small holes
layer to levels below those necessary to permit back-corona
in a similar rod at their midpoint, and threaded through holes
and sparkover, even with infinitely resistive particles. This
in a third rod at the upstream end. The spring tensions are
adjusted to reduce vibration of the corona wires due to effects
Paper IUSD 84-87, approved by the Electrostatic Processes Committee of of the corona wind and the turbulence.
the IEEE Industry Applications Society for presentation at the 1984 Industry
Applications Society Annual Meeting, Chicago, IL, September 30-October 4. Measurements of collection efficiency with both ac and dc
This work supported in part by Combustion Engineering Inc. and in part by excitations are made under conditions that are otherwise
the Cabot Solar Energy Fellowship Program. Manuscript released for identical. The ac excitation, shown in Fig. 1, consists of a
publication January 4, 1988.
R. M. Ehrlich is with Hewlett-Packard Laboratories, Building 24, 1501 variable-frequency oscillator and power amplifier driving an
Page Mill Road, Palo Alto, CA 94304. iron-core stepup transformer. Saturation of the latter sets the
J. R. Melcher is with the Department of Electrical Engineering and lower frequency limit at about 20 Hz. The rms electric field-
Computer Science, Massachusetts Institute of Technology, 77 Massachusetts
Avenue, Room 10-172, Cambridge, MA 02139. -
strength at the plates is typically Eo 3 x l o s V/m and the
IEEE Log Number 8820168. rms corona current density there is of the order of JO - 1 PA/

0093-9994/88/0700-0717$01.OO O 1988 IEEE

Authorized licensed use limited to: BTU Cottbus-Senftenberg. Downloaded on February 13,2024 at 20:48:53 UTC from IEEE Xplore. Restrictions apply.
718 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 24, NO. 4, JULYIAUGUST 1988

7
+ I
Weff. I

(rn's'
tI --
-I
+
+
+
++ j
t E
--++
- +
- +#+
+

V(kV)
aerosol
Fig. 2. Plot showing variation of effective migration velocity with applied
(dc) corona voltage. Measurements made using positive corona are labeled

-
" + " and those made using negative corona are labeled " - ".

aust
2. With equal magnitude voltages, both the corona current
density and the efficiency were higher with negative coronas.
Even with equal magnitude corona current densities, the
Y
corona
wires
I
spring
\
sam p l ing collection efficiency was higher with a negative corona. As
tube
has been reported in the past [4, p. 851, the negative dc corona
(b)
Fig. 1 . (a) Side view of laboratory-scale ac electrostatic precipitator. (b) Top
caused discrete white glowing points on the wires, separated
view of laboratory-scale ac electrostatic precipitator. by about 2 cm, while the positive dc corona gave off a more
uniform blueish glow. This is assumed to indicate a difference
cm2. A series capacitor at the high-voltage output of the in the relative spatial distributions of the positive and negative
transformer adds a positive bias to the voltage so that the time- dc coronas.
average corona current is zero. Shown in Fig. 3 is the dependence of weffon the rms voltage
The particulate used in these experiments was a monodis- with an ac excitation at 60 Hz and U = 4.3 m/s. The
perse aerosol of dioctyl phthalate (DOP) particles of 1-um frequency dependence of weff is shown in Fig. 4 for the same
-
diameter, with the number density, n 2.5 x lo9 m-3, at the mean flow velocity and an rms excitation voltage of 8 kV. The
solid curves in these last two figures are based on a theoretical
inlet of the channel. The efficiencies reported here were based
on measurements of the DOP mass loading at the outlet with model to be discussed in Section IV. Measurements of the
and without electrical excitation: efficiency were also made under nearly identical electrical
conditions but with U = 8.9 m/s: Expressed in terms of w,ff,
these observations were within experimental error of those
shown by Figs. 3 and 4. This encourages the presentation of
where manis the mass loading measured while the electrical the efficiency in terms of weff.
excitation is present and moffis the mass loading obtained with The experiments described above demonstrate the ability of
no electrical excitation. Thus the collection efficiencies ac corona excitations to precipitate particles from an air flow,
reported do not include mechanical collection. using zero time-average current. A more dramatic, although
artificial, comparison between dc and ac precipitation was
B. Experimental Results made by covering the collection plates with a thin layer of
Because it will be found to provide a useful correlation of insulating tape. In the case of dc excitations the corona current
the data, the effective migration velocity, weff,is used here to was completely extinguished and the collection efficiency was
represent the efficiency. According to the Deutsch (or approximately zero, regardless of the magnitude of the applied
complete mixing) model [4, pp. 164-1661, the efficiency, 7,is voltage. With ac excitations the precipitator functioned essen-
related to weffby tially as it had with the bare collection plates.
111. MODEL
7 = 1 - e-(L/W(d/weff)
= 1- e - ( 7 r / T m ) (2)
A . Problem Description
where d , L , and U are, respectively, half the channel height The situation is as pictured schematically in Fig. 5, which
and the channel length, and the mean air velocity, = L / U , depicts the upper half of a symmetric single-stage precipitator.
is the convection time of the air-flow, and T, = d/weffis the Air laden with particles flows between two parallel plates of
characteristic transit time for particles migrating across the length L and separation d. The lower plate takes the place of
channel. Here weffis defined as being related to the measured the wires in the center of an actual precipitator while the upper
efficiency by this expression solved for weff: plate represents a collection surface. The mean flow velocity
d U is assumed to be uniform in the core flow, decreasing
weff = - U - l n ( l - v ) . (3) rapidly to zero across thin boundary layers at each plate. The
L
particle concentration, n ( x , z , t ) , is taken as uniform at the
Typical effective migration velocities are shown as a inlet and is equal to no.
function of positive and negative dc excitation voltages in Fig. The particles are charged from an ac source at the mid-plane

Authorized licensed use limited to: BTU Cottbus-Senftenberg. Downloaded on February 13,2024 at 20:48:53 UTC from IEEE Xplore. Restrictions apply.
EHRLICH AND MELCHER: SINGLE-STAGE AC ELECTROSTATIC PRECIPITATION 719

n
Fig. 3. Plot showing variation of effective migration velocity with rms
magnitude of 60-Hz ac corona voltage.

Fig. 6. Sketch showing vertical displacement versus time for particle


migrating in applied precipitation electric field, in absence of turbulent
fluctuations in velocity of surrounding air.

It will be assumed that the device has been in operation for a


time sufficiently long for it to have reached a steady state
(except that the particulate accumulated on the upper wall
increases steadily with time). The efficiency of collection will
be computed in a manner that is consistent with the definition
f (Hr) given in (1):
Fig. 4 Plot showing variation of effective migration velocity with frequency
of 8-kV ac corona voltage.

Here ( n ( z ,L ) ) represents the time-averaged particle concen-


tration as a function of z at the exit of the precipitator (where x
= L).

B. Volume Equation
In the arguments below, the following abbreviations will be
. L P made where appropriate: the time-averaged particle concentra-
Fig. 5 . Idealized ac precipitation problem. tion ( n ) for ( n ( x , 2)); the fluctuating part of the particle
concentration n‘(x, z , t ) ; the fluctuating x component of gas
( z = 0) so that, were it not for the turbulent mixing present,
they would migrate along paths similar to that shown in Fig. 6. velocity u; for u; ( x , z, t ) ; uz’ for uz’ ( x , z , t ) ; and the
The oscillatory particle migrations which result from periodic oscillatory part of the particle migration velocity w ’ for
corona charge reversals were modeled by Ehrlich and Melcher w’(t).
[5], who characterized them by five quantities; T, the cycle As in the approach of Leonard, Mitchner, and Self [ 6 ] ,the
time of the electrical excitation; Imax,+ , the maximum particle instantaneous continuity equation for the particles is time-
excursion during a positive half-cycle of the excitation; I,,, + , averaged:
--
the net particle movement during the positive half-cycle; and d(n) a ( n ) an’u; an’ui a n “
the corresponding parameters for the negative half-cycle I, - U--+W,,,, -+-+- +-=o (5)
and I,,, - , respectively. In this work, symmetric charging will ax az ax az az
be assumed so that lmaX,+ = lmx,- = I, and lnet,+= lmt,- = where w,,, = lmt/T is the time-averaged particle-migration
I,,,,/2. In addition, it is assumed that the particles migrate only velocity.
a small distance during a single half-cycle so that l,, Q d. The last term in (3, which does not appear in the
The particle charging is assumed to have negligible effect on corresponding equation in the work, accounts for any correla-
the imposed electric field and ion flux densities. tion between the oscillatory particle migrations and the
The effects of turbulent mixing on particle transport in the fluctuations in particle concentration. The oscillatory particle
precipitator will be modeled by assignment of a uniform migrations are completely periodic with a maximum displace-
equivalent diffusivity, D t , in the bulk flow. A zone of ment l,,, 4 d. In the volume this term can be neglected on the
thickness, 6 4 d, in which no turbulent transport takes place, basis that there is no significant correlation between n ‘ and
is assumed to exist immediately adjacent to the plate at z = d. w ’ . The correlation between the fluctuating concentration of
The only way that particles can reach the collection plate from the particles and their oscillatory migrations may not be
the edge of the core flow is to migrate across this zone. Effects negligible at the collection plate as is argued in Section I11 C.
of re-entrainment are ignored so that once a particle reaches a Neglecting the last term of (5) in the volume and modeling
wall it has been “collected” and will remain at the wall. the effects of turbulent mixing with an equivalent turbulent

Authorized licensed use limited to: BTU Cottbus-Senftenberg. Downloaded on February 13,2024 at 20:48:53 UTC from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 24, NO. 4, JULYIAUGUST 1988

seen in Fig. 7(a) that a particle entering the sublayer when t =


0 would, if unobstructed by the wall, migrate a distance fnet/2
into the sublayer by the time t = t , . Outside the laminar
sublayer, turbulent diffusion maintains a particle concentration
n ( ( d - 6), x, t ) , which is approximately independent of time
[i.e., n'((d - 6), x, t ) Q ( n ( ( d - 6), x ) ) ] .
Because the sublayer is thin, particle conservation within is
Turbulent Flow represented by a conservation equation that neglects the x
(a) dependence. With w ( t ) defined as the z directed migration
velocity,
an a
-+- (wn)=O. (7)
at az
With the objective of evaluating the time-average migration
flux into the sublayer, (7) is integrated over the surface of Fig.
7(a) enclosed by the contour A -+B(O to tl in the t direction),
B C ( d - 6 to d - 6 + fne,/2in the z direction), and C
-+ -+

A along the particle trajectory. By the two-dimensional form


t=o liI2 t=T/2
Turbulent Flow of Gauss' theorem, this is converted into a contour integral:

:: a z
(b)
Fig. 7. (a) Paths taken by particles entering laminar sublayer, for case I,,,
- (Ine,/2)< 6. Dotted lines show paths particles would take if unobstructed
by collecting surface. (b) Paths taken by particles within laminar sublayer
s, [-+- ( w n ) ] dtdz= ( n c + wn;) &=O. (8)

for case I, - (Ine,/2)> 6.


Here, and are unit vectors in the t and z directions and d7
diffusivity result in [6]: is an incremental line element directed perpendicularly out-
ward from the contour. Under the assumption that along a
given particle trajectory n retains the same value as it had
when it entered the sublayer, (8) gives the number of particles
entering the sublayer per unit area in the case I,,,,, - lnet/2< 6
C. Boundary Conditions
as
The constraint on the inlet concentration n(0, z , t ) is as
stated in Section IIIA. The boundary condition on the
collection surface, the upper plate, is based on the principle
(n(d-6, x)) j: w d t = (n(x,d - 6 ) ) ic dz
that the only mechanism of particle transport across the
Let
laminar sublayer is migration. Here two distinct cases are = (n(x,d - 6 ) ) - (9)
examined; one in which lmax- lnet/2< 6, and the other in 2
which f,,, - fnet/2> 6. (In the case of an equality, the two
Equation (9) does not include a contribution from the contour
analyses give equivalent results.)
path CA because the integrand is normal to dTalong the path.
Fig. 7(a) shows a close-up of the region near the upper
A similar illustration for the case in which f,,, - lne1/2> 6
(collection) plate for the first case. The migration paths of
is shown in Fig. 7(b). In that figure, t2 is the earliest time at
particles entering the sublayer at various times during a single
which particles can enter the sublayer and still return to the
half-cycle of the electrical excitation are shown, with the
core flow. In this case the net flux of particles into the sublayer
vertical coordinate representing the particle displacement from
is evaluated by applying (8) to the contour ABCDA of Fig.
the wall and the horizontal coordinate representing time. The
+ 7(b). Because they are particle trajectories, the contributions
i, component of the velocity of particles within the laminar
on BC and DA are zero. Thus for f,,, - fne,/2 > 6,
sublayer varies from 0 at the upper plate to U at the edge of the
sublayer in a manner that depends upon the nature of the shear
flow there. All particles in the sublayer, however, share the
same E component of velocity w( t ) , independent of position
within the sublayer. = ( n ( x ,d-6))(fmax-6)* (10)
For convenience, the timescale begins with t = 0 at the Because 6 Q d , (9) and (10) can be evaluated using ( n ( d ,x ) )
moment the particles begin migrating toward the upper plate.
in place of ( n ( x , ( d - 6))) with little loss of accuracy.
The earliest time at which a particle can enter the laminar The time-average flux density of particles entering the
sublayer and still return to the turbulent core-flow is defined as
sublayer (I',) is given by dividing (9) and (10) by a half-
t , . All particles that enter the sublayer in the time interval [0, period:
t l ] will remain in the layer or at the wall at the end of the half-
cycle (at t = T/2). Any particles that enter during the time
interval [ tl , T/2] will return to the turbulent core flow. It is

Authorized licensed use limited to: BTU Cottbus-Senftenberg. Downloaded on February 13,2024 at 20:48:53 UTC from IEEE Xplore. Restrictions apply.
EHRLICH AND MELCHER: SINGLE-STAGE AC ELECTROSTATIC PRECIPITATION

where where 8, is the solution to the eigenvalue equation

--I
[ wp \/28m\
I - II - I

and tan (e,)=


[z ( 3 - 1 1 -(2)z
that lies in the range ( m - 1)a < Om < ma,
The max ( ) function in 12 selects the maximum of its
arguments. The quantity w, can be thought of as the
“precipitation velocity,” or the rate at which particles leave F, = 2 w [ ( l + w z [l+(~)’])”’-l]
the bulk flow at z = d .
Note that while l,, and I,,,,, can be set to the values predicted and PE wnetd/Dt. The electric Peclet number, P E ,can also
by Ehrlich and Melcher [5], the model presented here is be expressed as a ratio of time constants, PE = 7d/7,%4,where
applicable to any situation in which the particle motion is Td = d Z / D is
, the characteristic diffusion time based upon the
characterized by periodic and average parts (for example, in a half-width of the channel, and 7M = d/wnet.The following
two-stage precipitator with the second stage excited by a normalizations have been used in (16); z = z / d , x = x / d , and
biased sinusoidal voltage). In the case of a two-stage ac w = wnet/U.In the limits wP/wnet 00, wP/wnet= 1, the
-+
precipitator with no bias in the second-stage voltage, w,, = 0, eigenvalue equation, (17), approaches those given in (27) and
and so wp = wac.This is the precipitation velocity used by (29), respectively, of the work by Leonard, Mitchner, and
Bart, Melcher, and Ehrlich [7]. Self.
This constraint on particle precipitation can be expressed in By use of the orthogonality of the modes in (16), along with
terms of the concentration in the bulk flow by requiring that the boundary condition at the lower plate ((25) of Leonard’)
the flux of particles to the edge of the precipitator volume (due
to time-average migration and turbulent diffusion) be equal to
the precipitation rate specified by (1 1 ) . Thus

the values of the coefficients can be determined [8]:

The only targets for precipitation at the mid-plane ( z = 0)


of the precipitator are the wires, which present a small cross-
sectional area. In addition, the larger corona charge densities
in the vicinity of the wires will allow for quicker reversal of * I I
the particle charge than in the volume. For these reasons,
precipitation onto the wires will be neglected. Symmetry then
requires that the particle flux at z = 0 be zero:

[ %+
-D, w,.(n)] 1 z=o =O. where E , E 2e,/pE and g = 2( wP/wnet) - 1 are defined for
convenience. Together, (19) and (20) determine A, and B, ,
(20)

A downstream boundary condition on (n)can be obtained which can in turn be substituted into (16) for n ( z , x).
by assuming that the precipitator actually extends infinitely in Finally, the expression for n ( z , x) can be used in (4) to
the x direction (with the point x = L being defined for the determine the collection efficiency 7:
purpose of measuring the collection efficiency at a finite point)
2 e (1 12)PE
and that (n) + 0 as x + 00.
7=1--
PE
2 J(g2+Ei)(l+Ei)
(&?+I)
m=1
( - l)m+’

D . Problem Solution
Because of the similarity between this problem and the one
. Ame-FmWL/d (21)
9

addressed by Leonard, Mitchner, and Self [6], the modal The collection efficiency predicted by (21) can be expressed
expansion for the particle distribution is simply presented as a function of four dimensionless parameters. The represen-
here, without derivation. The solution to (6) satisfying the tation chosen here includes PE and g , as defined in the
boundary conditions of Section IIIC is preceding, along with K = 7 c / 7 M = w,,,L/dU, and y =
7d7M/(L/d)’7; = U2d/D,wnet.The electric Peclet number,
m
PE, characterizes the competition between particle diffusion
(n(z,x)) =no [A, cos (8,z)
e(1/2)PEz
m=I

’There is a typographical error in Equation (25) of Leonard’s paper.


+ B , sin ( 8 , ~ ) ] e - ~ m ~(16)
~ Equation (19) of this paper is a corrected version.

Authorized licensed use limited to: BTU Cottbus-Senftenberg. Downloaded on February 13,2024 at 20:48:53 UTC from IEEE Xplore. Restrictions apply.
722 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 24, NO. 4, JULYIAUGUST 1988

L I I I 1 1
and migration in the direction transverse to the flow. If PE 4 I.0C I

1, then the transverse particle distribution is dominated by


diffusion. The oscillatory particle migrations affect the effi-
ciency through the value of g. The parameter K characterizes
the competition between convection in the air flow, which
tends to take particles out of the precipitator volume via the
exit at the right in Fig. 5 , and migration, which tends to
remove particles via the collection plate at x = d. The Wp' W",t
Deutsch model for particle precipitation, (2), expresses the y = 1000

collection efficiency as a function of this parameter alone. The 0 I I I I


importance of axial diffusion is characterized by the value of 0 .8 1.6 2.4 3.2 4.O
K
y. If y % 1, then particle diffusion in the axial direction is Fig. 8. Plot showing predicted collection efficiency as function of K
unable to significantly alter the particle concentration gradi- (normalized electric field strength) for various values of PE(electric Peclet
ents caused by the particle convection and migration. The number), for case of wp =
parameter w = wnet/ U , which is used in (16), (18), and (2 l), I I I I I 1
can be expressed as w 2 = P E / while~ the term multiplying
-F, in (21) is seen to be equal to K.
Figs. 8-10 show the variation of the collection efficiency
predicted by the model with changes in P E , g, and K. For all
three illustrations the value of the fourth parameter is set at y
= 1000, a value that is typical for the experiments that have
i
been described. In Fig. 8 the precipitation velocity w, is equal
to the time-average particle migration velocity w,, so that the _I

situation there is much like that of a conventional (dc)


IY/ I
precipitator. Several curves, corresponding to various values ov
0
I
.4 .e
I
I. 2
I I
1.6
I
2.0
of PE, are shown. It is seen that with other parameters held K
fixed, larger turbulent diffusivities (corresponding to smaller Fig. 9. Plot showing predicted collection efficiency as a function of K
values of P E ) result in lower collection efficiencies. Physi- (normalized electric field strength) for various values of PE(electric Peclet
number), for case w, = 4wneI.
cally, the turbulence drives particles away from the collection
plate while the net migrations tend to accumulate them there. I .o
Similar sets of plots are shown in Figs. 9 and 10, for
sucessively larger values of wp relative to wnet.When w p > .e
wnetrthe oscillatory particle migrations contribute to the rate
of particle precipitation at the collection plate. It is seen that
for small values of K = r c / r M ,the collection efficiency
I
7
.6

actually improves with increasing turbulent diffusivity . Physi- .4


cally this is because the turbulence aids in transporting
particles to the vicinity of the collection plate. The particles 2

are precipitated from the edge of the flow region more rapidly
0
than their net migrations can bring them there. 0 .2 .4 .6 .8 I .o
K
Fig. 10. Plot showing predicted collection efficiency as function of K
IV. APPLICATION
OF MODEL (normalized electric field strength) for various values of PE (electric Peclet
number), for case of w, = OW,,^.
A . Discussion of Assumptions
The most questionable assumptions of Section I11 are those particles, 7p,due
to their net charge is
concerning the spatial uniformity of the electric field, ion-flux, 1 1
7 =--r eu
and the resulting particle migration velocity. The non-planar 18x
electrode geometry, ion space charge, and the finite ion transit where x 4/3.rrR3n is the volume fraction of the flow
time from the corona wires to the collection-plate (the ion occupied by the particles ( R is the particle radius), and rev
transit time is of the order of the half-cycle time of a 300-Hz
excitation) all contribute to particle charging and migration
~ l , , / e ~is i electroviscous time of the flow [vu 2 x
E the -
Kg m-l s-l is the viscosity of air). For the parameters of the
that varies both with time and space. However, the simplifica- -
experiments of Section 11, rp 1000 s, much greater than the
tion afforded by the assumption of uniform electric field and
ion-flux density allows attention to be focused upon the
convection time, rc -
1 s, of the flow in the channel. Thus
particle space charge did not have an appreciable effect on the
interactions between diffusion, migration, and convection in collection efficiency.
the precipitator. More sophisticated models of the ac charging
and migration processes could include the effects neglected B. Further Simplifications
here. Although the model of Section I11 is useful for predicting the
Ehrlich [3] showed that the self-precipitation time of the effects of the finite turbulent diffusivity , oscillatory particle

Authorized licensed use limited to: BTU Cottbus-Senftenberg. Downloaded on February 13,2024 at 20:48:53 UTC from IEEE Xplore. Restrictions apply.
~

EHRLICH AND MELCHER: SINGLE-STAGE AC ELECTROSTATIC PRECIPITATION 723

migrations, and laminar sublayer thickness on the collection well. Ehrlich and Melcher [5] showed experimentally that the
efficiency, it contains more complexity than is necessary in ac particle charging and migration equations used here to
many limiting situations. It is shown here that for the limit of predict w,,, and w, are not accurate for frequencies above
large turbulent diffusivity and a thick laminar sublayer, the about 150 Hz. In addition, the model presented here cannot
model of Section 111 reduces to the familiar Deutsch model for account for the substantial differences between the collection
particle precipitation from a turbulent flow. This simplified efficiencies obtained with positive and negative dc coronas of
model is then applied to the experimental situation of Section similar magnitude. Factors neglected by the ac charging model
11. include 1) nonuniform electric-field and corona ion charge
From (12) and (13) and the definition of wnet Inet/Tit is densities, 2) finite ion transit times, and 3) differences between
seen that wp = wnetrunless I,,, - (lne1/2)> 6. Bart, Melcher, the spatial distributions of the positive and negative coronas.
and Ehrlich [7] showed that with flow parameters similar to
those used here, the laminar sublayer thickness 6 was of the B. Practical Applications
order of or greater than 160 pm. Using equations derived by
Ehrlich and Melcher [5] for I,,, and I,, along with estimated The performance of the ac precipitator with source frequen-
parameters of the experiments reported here, Ehrlich [3] cies in the range f -
50 - 60 Hz is of primary interest
showed that l,,, - (lne,/2) < 20 pm. Thus, with the
because of the relative ease of obtaining such excitations. In
oscillatory particle migrations unlikely to play any significant the laboratory-scale precipitator described here, the effective
role in determining the particle precipitation rate, it can be particle migration velocity obtained with a 60-Hz ac excitation
assumed that wp = wnetso that g = 1. of a given rms voltage magnitude was about 28 percent of that
In the limit y >> 1 >> PEthe first mode in (16) dominates the measured using a negative dc excitation of the same voltage
particle concentration distribution2. In that limit, O1 + level. If this were assumed to hold true for commercial
6,
A , + 1 , F, + 1 , and A,,, 0 for m = 2, 3,
+ The devices, then under ideal conditions (low particle resistivity)
- . e .

collection efficiency is an ac precipitator would have to be about four times as large as


its dc counterpart to attain the same collection efficiency. As
the particulate resistivity rose, however, the performance of
the dc device would be degraded while the ac precipitator
Thus for large values of the turbulent diffusivity D , , the would presumably continue to operate at roughly the same
efficiency can be predicted by a Deutsch model. Ehrlich and efficiency. Thus a parameter of some important is the
Melcher [9] made measurements of the effective particle particulate resistivity at which the effective migration veloci-
diffusivity in the axial direction under circumstances similar to ties obtained with the dc and ac precipitators are equal.
those used here, with U = 4.3 m/s, D, 8 x - m2/s. If it White [ 101 studied the effects of high particle resistivity on
is assumed that the longitudinal diffusivity is of the same the performance of conventional (i.e., dc) precipitators. He
order, then values for PEand y can be computed using U and found that the effective migration velocity fell by about a
-
Dt, along with d = I .5 cm, and w,,, 10 cm/s, and are seen factor of 4 as the resistivity rose in the range p 3 x lo9 -+ 7 -
to satisfy the given requirement. x 10" Q-cm. From this, one might infer that ac precipitators
could out-perform their dc counterparts with particulate
C. Comparison to Data resistivities in excess of p 1012 Q.cm -
The solid lines in Figs. 3 and 4 were obtained by setting weff The corona current density used JO -
1 pA/cm2 was about
= wnet as computed by the ac charging model [5]. The two an order of magnitude higher than those typically found in
parameters necessary to compute w,,, as a function of conventional precipitators, which resulted in a particle-
frequency for a given rms excitation voltage were wo,the charging time constant Tch -
3 ms that was about an order of
effective particle migration velocity with positive dc corona magnitude shorter than normal. Such a large current-density
charging at the same voltage magnitude, and Tch , the charging was necessary to insure sufficient particle charge levels with a
time. For each voltage used in the experiment, wo was 60-Hz excitation. If the corona frequency were lower, then the
obtained from dc precipitation data (Fig. 2). The value used current density could be correspondingly smaller.
for Tch was 3 ms, which was consistent with the corona ion
current densities used. C. Unresolved Issues

V. CONCLUSION The major motivation for the use of ac precipitators is the


hope that the ac excitation will either prevent the occurrence of
A . Model Accuracy back corona or lessen its effect on the collection efficiency.
The simplified model of Sections 111 and IV is seen (in Figs. While the experiments and models described here demonstrate
3 and 4) to predict the performance of the ac precipitator as a the ability of an ac precipitator to collect particles (even with
function of frequency and rms voltage magnitude. It is the collection electrodes covered by insulation), they do not
tempting, therefore, to assume that it accurately models the provide any conclusive evidence that such a precipitator will
mechanisms of particle charging, migration, and turbulent actually collect highly resistive particles with a greater
transport that contribute to the overall collection efficiency as efficiency than that obtained with a conventional device.
Experiments should be performed under conditions that more
* The limit on y is necessary to insure that particle diffusion in the axial closely simulate an industrial application. Specifically, highly
direction is negligible. resistive particulate should be used over a long enough period

Authorized licensed use limited to: BTU Cottbus-Senftenberg. Downloaded on February 13,2024 at 20:48:53 UTC from IEEE Xplore. Restrictions apply.
724 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 24, NO. 4, JULYIAUGUST 1988

of time to allow a study of the occurrence of back corona as Richard M. Ehrlich (M’86) was born in Chicago,
well as its effects on the collection process. The adhesion of IL. He received the B.S., the M.S., and the Sc.D.
degrees in electrical engineering from the Massa-
the collected layer should be studied to determine whether the chusetts Institute of Technology, Cambridge, MA,
alternating fields will result in an increased re-entrainment in 1977, 1979, and 1984, respectively.
problem. Since May 1984 he has been a Member of the
Technical Staff at Hewlett Packard Laboratories,
ACKNOWLEDGMENT where he works on servo-control issues associated
with magnetic disc drives.
This work formed part of a thesis submitted by the first Dr. Ehrlich is a member of the IEEE Magnetics
author to the Dept. of Electrical Engineering and Computer and Control Systems Society
Science at M.I.T. in partial fulfillment of the requirements for
the Sc.D. degree, with Gerald Driggers of Combustion
Engineering, Inc. both administering the project and function-
ing informally as a thesis committee member. Formal mem-
bers of the committee were J. Haritonidis, L. Smullin, M.
Zahn, and the second author.

REFERENCES James R. Melcher (S’56-M’62-SM’75-F’77) re-


H. Lau, “AC electrostatic precipitator,” Staub-Reinhalt Luft., vol. ceived the B.S degree in electrical engineering and
29, no. 8, pp. 10-12, Aug. 1969. the M.S. degree in nuclear engineering, both from
H. Krug, “Physical principles of ac-operated electrostatic precipita- Iowa State University, Ames, and the Ph.D. degree
tors,” Luftverunreinigung, vol. 40, pp. 40-46, Oct. 1969. in electrical engineering from the Massachusetts
R. M. Ehrlich, “AC electrostatic precipitation,” Sc.D. thesis, Dept. of Institute of Technology, Cambridge.
Elect. Eng. and Comp. Sci., M.I.T., Cambridge, MA. Since joining the M.I.T. faculty in 1962, his
H. J. White, Industrial Electrostatic Precipitation. Reading, MA: research and teaching activities have been in the
Addison-Wesley, 1963. broadly defined area of continuum electrome-
R. M. Ehrlich and J. R. Melcher, “AC corona charging of particles,” chanics. A continuing theme has been the develop-
IEEE Trans. Ind. Appl., vol. IA-23, no. I , pp. 103-107, Jan./Feb. ment of electrohydrodynamics as an engineering
1987. science. Applications of his research typically involve the effects of electric
G . Leonard, M. Mitchner, and S. A. Self, “Particle transport in and magnetic fields on processes of energy conversion, on the transfer of
electrostatic precipitators,” Atmospheric Environ., vol. 14, pp. mass, momentum, or heat, and on the control of media. Typical areas of
1289- 1299, 1980. application are air pollution control, power systems, and manufacturing
S. F. Bart, J. R. Melcher, and R. M. Ehrlich, “AC collection processing. He is the Stratton Professor of Electrical Engineering and Director
efficiency of pre-charged particulate in turbulent flows,” Ind. Eng. of the Laboratory for Electromagnetic and Electronic Systems.
Chem. Res., no. I , pp. 123-131, 1988. Dr. Melcher is the author of more than 90 journal publications, the
F. B. Hildebrand, Advanced Calculus f o r Applications. Englewood, monograph Field-Coupled Surface Waves, the text Continuum Electrome-
NJ: Prentice-Hall, 1962, p. 206. chanics, and coauthor of Electromechanical Dynamics and Electromag-
R. M. Ehrlich and J. R. Melcher, “Turbulent diffusion of a travelling netic Fields and Energy. In addition, he has made three commercially
wave of charged aerosol,” Ind. Eng. Chem. Res., vol. 26, no. 3, pp. available educational films. He is a fellow of the IEEE, a member of the
4 5 6 4 6 2 , 1987. National Academy of Engineering, and in 1971-1972 he held a Guggenheim
H. J. White, “Resistivity problems in electrostatic precipitation,” J. fellowship while a Fellow of Churchill College, Cambridge University,
Air Pollution Cont. SOC.,vol. 24, no. 314, 1974. Cambridge, England.

Authorized licensed use limited to: BTU Cottbus-Senftenberg. Downloaded on February 13,2024 at 20:48:53 UTC from IEEE Xplore. Restrictions apply.

You might also like