EPA-600/2-78-032
February 1978
Environmental Protection Technology Series
                                         EVALUATION
                             OF THREE  INDUSTRIAL
                        PARTICULATE SCRUBBERS
                              Industrial Environmental Research Laboratory
                                    Office of Research and Development
                                   U.S. Environmental Protection Agency
                              Research Triangle Park, North Carolina 27711

-------
                                                  EPA-600/2-78-032
                                                      February 1978
             EVALUATION OF  THREE
INDUSTRIAL PARTICULATE SCRUBBERS
                              by

                     Seymour Calvert, Harry F. Barbarika,
                        and Gary M. Monahan

                       Air Pollution Technology, Inc.
                     4901 Morena Boulevard, Suite 402
                       San Diego, California 92117
                        Contract No. 68-02-1869
                      Program Element No. 1AB012
                         ROAP No. 21ADM-029
                     EPA Project Officer: Dale L Harmon

                  Industrial Environmental Research Laboratory
                    Office of Energy, Minerals, and Industry
                     Research Triangle Park, N.C. 27711
                           Prepared for

                   U.S. ENVIRONMENTAL PROTECTION AGENCY
                     Office of Research and Development
                        Washington, D.C. 20460

-------
                  RESEARCH REPORTING SERIES


Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination  of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series pre:

    1. Environmental Health Effects Research

    2. Environmental Protection Technology

    3. Ecological Research

    4. Environmental Monitoring

    5. Soct'oeconomic Environmental  Studies

    6. Scientific and Technical Assessment Reports (STAR)

    7. Interagency Energy-Environment Research and Development

    8. "Special" Reports

    9. Miscellaneous Reports

This report has been assigned to the  ENVIRONMENTAL. PROTECTION TECH-
NOLOGY series.  This series describes research performed to develop and dem-
onstrate instrumentation, equipment,  and methodology to repair or prevent en-
vironmental degradation from point and non-point sources of pollution. This work
provides the new or improved technology required for the control and treatment
of pollution sources to meet environmental quality standards,
                       EPA BEVIEW NOTICE


This report has been reviewed by the U.S. Environmental Protection Agency, and
approved for publication. Approval does not signify that the contents necessarily
reflect the views and policy of the Agency, nor does mention of trade names or
commercial products constitute endorsement or recommendation for use.

This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.

-------
                           ABSTRACT

     Field measurements on three full scale industrial scrubbers
were carried out to determine scrubber performance characteris-
tics, including the particle collection efficiency as a function
of particle diameter.  The three scrubbers were different gas-
atomized spray types with pressure drops ranging from 54 to 178
cm W.C.  Their performance on major sources of fine particle
emissions was compared to a mathematical performance model for
venturi scrubbers.
                               111

-------
                           CONTENTS

                                                              • * •
Abstract .......... , .............  •  •  11:L

Figures ......... ..................  v

Tables .............  i  .............  vii

Abbreviations and Symbols.  ...  ..............  i-x

Acknowledgment .......................  X1
    1. Introduction

    2. Conclusions
    3. National Dust Collector Model 850 Variable Rod
       Module Venturi .................... 4

    4. American Air Filter Kinpactor Venturi ........ 16

    5. Gas-Atomized Spray Scrubber ............. 31

    6. Performance Comparison ................ 47

    7. Performance Test Method ............... 64

    8. Data Reduction and Computation Method ........ 74


References .................. ....... 85
                                 IV

-------
                          FIGURES

Number                                                    Page

  1   Schematic flow diagram of scrubbing system 	  5

  2   Schematic of scrubber tower. 	  6

  3   Penetration versus aerodynamic particle diameter for
      ductile operation	 12

  4   Penetration versus aerodynamic particle diameter for
      gray operation	12

  5   Schematic diagram of scrubbing system	17

  6   Particle penetration for run 2	24

  7   Particle penetration for runs 4 and 5	24

  8   Particle penetration for runs 6, 7 and 8	25

  9   Particle penetration for runs 9, 10 and 11	25

 10   Particle penetrations for runs 12 and 13	26

 11   Schematic drawing of scrubber system ........ 32

 12   Particle penetration for cascade impactor runs 1
      and 2	41

 13   Particle penetration for cascade impactor runs 3
      and 4	41

 14   Particle penetration for cascade impactor runs 5,
      6 and 7 and diffusion battery runs 3-9 (inlet) and
      10-12 (outlet)	42

 15   Particle penetration for cascade impactor run 8. . . 42

 16   Particle penetration for cascade impactor runs 9
      and 10	43

 17   Particle penetration for cascade impactor runs 11,
      12 and 13, and diffusion battery runs 15 and 16
      (inlet)  and 13 and 14 (outlet)	43

-------
                       FIGURES (continued)

Number                                                       Pa&e

  18   Particle penetration for cascade impactor runs 14
  28
       and 15,
                                                               44
  19   Comparison of predicted and measured penetration for
       the NDC venturi during gray iron operation	51

  20   Comparison of predicted and measured penetration for
       the NDC venturi during ductile  iron operation 	 51

  21   Comparison of predicted and measured penetration for
       the average of runs  12 and  13 of the AAF  venturi.  ... 54

  22   Comparison of predicted with measured penetration for
       average of runs 5, 6 and 7,  for  the gas-atomized
       scrubber	59

  23   Comparison of predicted with measured penetration for
       run 8 for the gas-atomized  scrubber	59

  24   Comparison of predicted with measured penetration  for
       average of runs 9 and 10 for the  gas-atomized scrubber. 60

  25   Comparison of predicted with measured penetration  for
       average of runs 11,  12 and  13 for the gas-atomized
       scrubber	60

  26   Comparison of predicted with measured penetration  for
       average of runs 14 and 15 for the gas-atomized
       scrubber.	61

  27   Modified EPA sampling  train  with  in-stack  cascade
       impactor	    67
Schematic diagram of diffusion  battery system 	 69
                                VI

-------
                             TABLES
Number                                                     Page
   1   Process conditions	   8
   2   Inlet and outlet size distribution summary using log-
       probability analysis	10
   3   Diffusion battery size distribution analysis	11
   4   Overall penetration summary 	 13
   5   Opacity summary 	 14
   6   Cost data	15
   7   Inlet process conditions	19
   8   Outlet process conditions 	 19
   9   Outlet average gas composition	20
  10   Particle size distribution summary	22
  11   Mass loading and overall penetration	27
  12   Opacity	28
  13   Scrubber conditions 	 34
  14   Process conditions	34
  15   Average gas composition	36
  16   Cascade impactor data using log-probability analysis. 38
  17   Diffusion battery particle size distributions .  ... 38
  18   Mass loading and overall penetration	40
  19   Conditions for variable rod venturi performance
       predictions	50
  20   Conditions for AAF venturi performance prediction .  . 53
                                 VII

-------
                        TABLES (continued)

Number                                                     Page

  21   Pressure drop comparison for the gas-atomized spray
       scrubber	b8

  22   Conditions for gas-atomized spray scrubber perfor-
       mance prediction	5°

  23   Overall penetration comparison for the  gas-atomized
       spray scrubber	"2

  24   Measuring equipment and methods	65
                              Vlll

-------
                  LIST OF  ABBREVIATIONS AND SYMBOLS
  A,B   =  parameters  in equation (21)  or equation (29),
         dimension!ess
    b   =  Weibull slope,  dimensionless
    B   =  venturi parameter (equation  5),  dimensionless
   Cf   =  Cunningham slip correction factor (equation 9),
         dimensionless
  CDF   =  cumulative  distribution function, dimensionless
  C~   =  drag coefficient at throat inlet, dimensionless
   C   =  cumulative  particle mass concentration, mg/DNm3
 C t-   =  total inlet particle mass concentration, mg/DNm3
 C     =  total outlet particle mass concentration,  mg/DNm3
    d"  =  differential operator
   d   =  wire diameter of screen in screen diffusion battery,  cm
    w
   d,  = drop diameter,  cm or ym
   d.   = jet diameter,  cm
   d   = particle diameter,  cm
    P
  d    = particle aerodynamic diameter,  ymA
   pa
 d     = impactor stage cut  diameter,  ymA
  pciC
  d    = diffusion battery stage cut diameter,  cm
  d    = particle geometric  mass mean aerodynamic diameter,  ymA
   Jr o
  d    = particle mass  median aerodynamic diameter,  ymA
  d    = particle geometric  number (count) mean diameter,  ym
  d    = Weibull minimum particle diameter, ymA
f(d )   = particle frequency  distribution, dimensionless
   K   = inertial parameter  equation (6) or equation (7),
    P     dimensionless
   £.   = venturi throat length,  cm
   m   = mass of particles over  a differential  size element, g
   N   = cumulative number concentration of particles,  #/cm3
    P   = cumulative mass fraction (equation 13),  dimensionless
   Pt   = particle penetration, fraction
   Ptf  = overall penetration, fraction or %
   Qr   = gas volume flow rate, cm3/s
   QT   = liquid volume  flow  rate, cm3/s
    Li
                                  ix

-------
         LIST OF SYMBOLS AND ABBREVIATIONS  (continued)
  S  = solidity factor of screen diffusion battery stage,
       dimensionless
u, * = ratio of velocity attained by the liquid drops at the
       throat exit to the gas velocity, dimensionless
uGt  = ^as velocity i-n throat, cm/s
 u-  = jet velocity, cm/s
 u   = superficial gas velocity, cm/s
  j
  x  = length parameter (equation 3), dimensionless
  x  = parameter defined in equation (14), dimensionless
x,y  = variables in equation (21), dimensionless
GREEK
 AP  = pressure drop, cm W.C.
  0  = Weibull characteristic particle diameter, ymA
 PQ  = gas viscosity, poise (g/cm-s)
ymA  = aerodynamic micrometers (equation 7)
 p   = particle density, g/cm3
 PG  = gas density, g/cm3
 PT   = liquid density, g/cm3
 CT   = geometric standard deviation,  dimensionless
  o

-------
                         ACKNOWLEDGEMENT
     A.P.T., Inc. wishes to express its appreciation to Mr.  Dale
L. Harmon, the EPA project officer, for excellent coordination
and assistance in support of our technical effort.   The cooper-
ation provided by the plant personnel at the sites tested is
also greatly appreciated.
                                 XI

-------
                           SECTION 1
                          INTRODUCTION

     Air Pollution Technology, Inc.  (A.P.T.) conducted performance
evaluations on three industrial particulate scrubbers under EPA
Contract No. 68-02-1869.  The objective of the performance evalu-
ations was to determine fine particle penetration as a function
of particle size and scrubber parameters.
     The project involved the following tasks:
     1.  Locate suitable scrubbers for evaluation and select
those which:
         A. Have potential for high collection efficiency of
            fine particles,
         B. Control a process which is a major source of fine
            particulate emissions,
         C. Are widely used in industry.
     2.  Conduct appropriate field test programs to obtain neces-
sary performance data.
     3.  Use data obtained to evaluate the performance and pro-
bable economics of the scrubber.
     4.  Develop a useful performance model of the scrubber system,
     The first task involved an extensive correspondence campaign.
Many leads were taken from listings in the National Emissions Data
System (NEDS) and contacts with state and local air pollution con-
trol districts.  The result of this task was the identification of
the three scrubbers whose evaluations are presented in this report,
     A separate detailed report has been issued for each of the
scrubbers tested (Calvert, et al.  1976, Calvert, et al.  1977a>
and 1977b).  In this report each scrubber is discussed in a sepa-
rate section and comparison with a performance model is  made in a
succeeding section.  Also included in this report are discussions
of the testing and the data reduction methods that were used.

-------
                            SECTION  2
                          CONCLUSIONS

     Field measurements on three full scale industrial scrubbers
were conducted to determine scrubber performance characteristics,
including the particle collection efficiency as a function of
particle diameter.  A summary of the performance tests is given
in the tabulation below:
                                                 Pressure Drop
Control Device                Source                cm W.C.
Variable Rod Venturi        Iron Cupola               178
(National Dust Collector)
Venturi                  Borax Fusing Furnace         110
(American Air Filter)
Gas-atomized Spray          Iron Cupola              54-104
Scrubber

     All three scrubbers were in the general classification of
gas atomized spray scrubbers and their performance was compared
to the mathematical model of Yung,  et al.  (1977a).
     The program achieved the primary objective of evaluating
full scale industrial scrubbers, however some difficulty was en-
countered in finding suitable and willing  facilities to evaluate.
The following conclusions regarding venturi performance are
based on the results of our evaluations.
     1.  Comparison of the measured particle collection efficiency
to the mathematical model led to the following conclusions:
        (a)  The comparison was good in the aerodynamic particle
            diameter region around 1 ymA.
        (b)  In all cases the model predicted lower efficiency for
            particle diameters less than 0.5 umA. To some  extent
            this  is  expected because the model considers only in-
            ertial  impaction as  the  collection mechanism.

-------
        (c)  The comparison was not always good for particle dia-
            meters greater than 2 \irnK.  The model prediction in
            that size region was quite sensitive to the liquid-
            to-gas ratio, which usually was not known with pre-
            cision.  Thus the model may have been adequate in
            the range above 2 ymA, however knowledge of important
            variables was not precise and the adequacy of the
            model could not be fully proven.
        (d)  Predictions of overall efficiency were usually poor.
            The model, which assumed a log-normal size distri-
            bution, consistently underpredicted efficiency.  Two
            probable reasons are that the inlet distribution had
            a geometric mass mean diameter greater than was mea-
            sured with the heated cascade impactors and the size
            distribution was not log-normal.
                 Generally the model and the measured performance
            agreed.  However, each source and scrubber had parti-
            cular characteristics which requireed that some ana-
            lysis be made to reconcile the model with the data.
            Therefore, the model should only be used with a full
            awareness of the assumptions on which it is based.
            The model may need modification if used in situations
            where conditions are significantly different from the
            assumptions in the model.
     2.  High pressure drop scrubbers are vulnerable to leaks which
decrease the efficiency and waste power.  Leaks generally develop
because of improper construction materials and/or inadequate main-
tenance.
     3.  The cyclone entrainment separators used on the venturi
and the gas-atomized spray scrubber were inefficient and caused
both emissions problems and blower maintenance problems.

-------
                           SECTION 3
               NATIONAL DUST COLLECTOR MODEL 850
                 VARIABLE ROD MODULE VENTURI

 SOURCE AND CONTROL SYSTEM
     The  emission source was an iron cupola which operated for
 ten  hours per day.  The cupola melted both the relatively clean
 ductile iron and the dirtier gray iron.
     The  primary component of the scrubbing system was a National
 Dust Collector Model 850 variable rod module venturi scrubber.
 The nominal capacity was 10.4 m3/s at 21°C (22,000 CFM at 70°F).
 The variable rod module could be operated at pressure drops from
 100 to 230 cm W.C. (40 to 90 inches W.C.).  The venturi module
 consisted primarily of several parallel rods positioned normal
 to the gas flow.  The gas was moved through the scrubber by two
 induced draft Buffalo No. 91 blowers used in series.  Each blower,
 run by a  261 kW (350 HP) motor, was capable of drawing 114 cm
 W.C. (45  inches W.C.).
     The overall scrubbing system is shown in Figure  1 .  Gas
 flow from the cupola is controlled by adjusting the combustion
 (tuyere)  air flow.  The charge door is open so air continually
 leaks into the cupola.  The temperature at the top of the cupola
 is only about 175-205°C instead of the much higher temperature
 encountered with most cupolas.
     A series of quench sprays are introduced at the entrance of
 the connecting duct between the cupola tower and the scrubber.
About 5.8 £/s of fresh water is used in the quench section.  The
 temperature of the gas is reduced to about 35°C by the quench
 sprays.   The gas is saturated when it reaches the scrubber.
     As  shown in Figure  2 , the scrubber consists of the follow-
 ing sections:
     1.  A pre-spray section which sprays scrubbing liquor into
 the gas  stream at the rate of 7.8 £/s.  This pre-spray has slight
                                 4

-------
    TO
ATMOSPHERE
     t
S
•H
e
   O I
   to r^
   rj LO
         STACK
         OUTLET
         SAMPLING
           PORT
                                   t
                                VARIABLE
                                  ROD
                                 MODULE
                                VENTURI
                                SCRUBBER
                                                           V
 c
•H
--. u
m o
 S i/>
< to
o
o
                                                        QUENCH
                                                         WATER
                                                       350£/min
                                                          16°C
       INLET
       SAMPLIN
      I PORT
                                                        CUPOLA
                                                    TUYERE AIR
   I.D.  BLOWERS
                                                                     o
                                                                     CN1
                                                                     LO
                                                                    z
                                                                           AIR

             Figure 1.  Schematic flow diagram  of  scrubbing system.

-------
GAS OUTLET
  18-21°C
                                              COOLING WATER
                                                 SPRAYS
         COOLING
  SEPARATOR DECK
 COLLECTOR DECK
   VARIABLE ROD
      MODULE
                   •.•.•.is!/-:- •«••. •.:•.•.-.•/.v ••/*%» v.VJ'Wid
          DECKS   ••.••:•.•.•••/••• :•;«%•..-..•.•.:.•. .•."••..•;•:
 COOLING  WATER
    RETURN

SCRUBBER  LIQUOR
   RETURN
                                                 SCRUBBER  LIQUOR
                                                     SPRAY

                                                   PRE-SCRUBBER
                                                   LIQUOR  SPRAY
                                                      GAS  INLET
                                                      1,010 Am3/min
                                                      35 C
                   QUENCH/SCRUBBER LIQUOR
                            RETURN

          Figure 2.  Schematic of scrubber tower.

                                    6

-------
cooling effect, which may cause  some condensation growth of the
particles, and it may collect some particles as it settles out
onto the bottom of the scrubber  tower.  The drops from these
sprays are not carried through to the variable rod module.  The
liquor returned from this section is sent to the sludge tank.
     2. The high energy section, which  consists of a scrubber
liquor spray of 28.3 £/s and the variable rod module is next.
The pressure drop across the variable rod module is about 178
cm W.C. when the total pressure  drop across the whole tower is
224 cm W.C.  The temperature of  the scrubber liquor is about 35°C
and the ratio of liquid to gas volume flow rates in this section
is about 1.68 £/m3.
     3. The collector deck, consisting  of 7.6 cm of loose 2.5 cm
diameter marbles, separates out  the entrained liquid from the
high energy section.  The returned scrubbing liquor is split
into streams, with 801 going to  the recycle tank and 20% going to
the sludge tank.
     4. Above the collector deck is a plate which collects the
water from the cooling section.
     5. The gas is cooled prior  to leaving the scrubber tower
in a cooling section consisting  of sprays and two decks of single
layer 2.5 cm diameter marbles.   The cooling water is supplied at
about 15°C at the rate of 112 £/s to cool the gas to about
18-21°C.  There is no carryover of entrainment from the cooling
section.

-------
PROCESS CONDITIONS
     The operating conditions of the variable rod venturi scrubber
for the period of sampling are shown below:
                   TABLE I.  PROCESS CONDITIONS
   CONDITION
  INLET SAMPLE PORT
OUTLET SAMPLE PORT (1)
   Temperature
   Velocity
   Am3/s
   ACFM
   DNm3/s @ 0°C
   DSCFM § 21°C
   Vol % H20 Vapor
   Static Pressure
       32-35°C
9.2 m/s (30.3 fps)
        16.8
       35,700
        12.9
       27,300
        6.6
     -1.1 cm W.C.
     57-66°C
25.3 m/s (83.1 fps)
       20.5
     43,500
       14.5
     30,800
      3.3
 +0.38 cm W.C.
   Pressure Drop Across Tower
   Pressure Drop Across Venturi Module
   Superficial velocity Through
   Scrubber Tower (diameter 3.12 m)
   based on inlet ACFM
                    224 cm W.C
                    178 cm W.C,
                    2,19 m/s
   (1)   The outlet  sample port  was  located in the exit stack
   after the straightening vane section which was used to reduce
   the  tangential spin  of the gas.   For this reason outlet flow
   rates may not be as  reliable as  inlet flow rates.

-------
CASCADE IMPACTOR DATA
     Sets of data were obtained from the variable rod module
venturi scrubber as will be described later in the test method
section.  Fully developed flow enabled representative one-point
sampling.  Identical one-point sample locations were used for all
the data points.
     A summary of the inlet and outlet size distribution tests
is given in Table 2.  The distributions approached log-normality
for diameters between 0.3 and 2 umA.
     Average sample times for the inlet were three to four minutes
depending on the mass loading, while the outlet sample times
averaged approximately forty-five minutes to one hour.
     No inlet data were obtained for run 1 due to the excessive
vacuum at the start of the run which overloaded the impactor.
     Runs 9, 10, and 11 were not used as part of the data set be-
cause two of the five spray nozzles in the scrubber liquor spray
section were disconnected while these runs were being made.
     The data from the inlet cascade impactor runs was analyzed
to determine the effect heating the impactors above the stack
temperature had on the size distributions.  We could see no de-
finite trend in the heated impactor runs toward smaller size dis-
tributions than the unheated impactor runs.
DIFFUSION BATTERY DATA
     Diffusion battery data were taken during the testing period.
The runs were made alternately on inlet and outlet sample loca-
tions as shown in Table 3, while impactor runs were being per-
formed .
     Since operation of the scrubber was fairly constant over the
testing period the inlet and outlet samples were averaged result-
ing in one set of data each for the ductile and gray operations.

-------
   TABLE  2 .  INLET AND OUTLET SIZE DISTRIBUTION SUMMARY
              USING LOG-PROBABILITY ANALYSIS
RUN        OPERATION            INLET           OUTLET

N0-                         dpg,ymA  ag      dpg,ymA
1
2
3
4(D
5
6(2)
7(3)
8(4)
9(5)
10
11
12
IS^
14
15(7)
16
16P
17
Ductile
Ductile
Ductile
Ductile
Ductile
Gray
Gray
Gray
Gray
Gray
Gray
Gray
Gray
Gray
Ductile
Ductile
Ductile
Ductile
_ _
0.69
0.76
0.25
0.27
1.10
1.51
0.56
0.54
0.62
1.90
0.62
0.59
0.72
0.84
0.58
0.71
0.49
—
2.0
2.0
1.6
1.5
1.7
1.8
1.6
1.5
1.6
1.6
1.6
1.6
1.6
1.6
1.5
1.7
1.8
0.48
0.52
0.52
0.52
0.56
0.58
0.56
0.52
0.63
0.63
0.81
0.51
0.51
0.52
0.50
0.48
0.48
0.49
1.7
1.8
1.9
1.8
1.8
1.7
1.8
1.6
1.6
1.7
1.7
1.6
1.6
1.5
1.7
1.7
1.7
1.7
(1)   Run 4 had 40% of the 5th stage holes plugged on the
     inlet sample
(2)   Run 6 had 30% of the 3rd stage holes plugged on the
     inlet sample
(3)   Run 7 had 30% of the 6th stage holes plugged on the
     inlet sample
(4)   Run 8 had 20% of the 6th stage holes plugged on the
     inlet sample
(5)   Run 9 had 30% of the 6th stage holes plugged on the
     inlet sample
(61   Run 13 had 4 holes  of the 5th and 6th stages plugged
     on the inlet sample
(7)   Run 15 had 50% of the 4th stage holes plugged on the
     inlet sample
                               10

-------
      TABLE 3.  DIFFUSION BATTERY SIZE DISTRIBUTION ANALYSIS

Operation

Ductile
Gray
Inlet
d ~ ,,™ a
pn,ym g
0.023 3.0
0.040 3.1

d
_e
0.
0.
Outlet

n, ym
070
090

0
_!
2.2
2.0
PARTICLE PENETRATIONS
     Particle penetration versus particle aerodynamic diameter
was computed and is shown in Figures 3 and 4 for ductile and gray
iron operations respectively.  Penetrations for a few of the runs
are not shown because either their size distributions or their
total loadings were outside a standard deviation from the mean
of the set of data within which the run belonged.
     The penetrations were calculated using a mathematical for-
mula based on the log-normality of the inlet and outlet size dis-
tributions .
     Diffusion battery data yield penetration related to physical
size while cascade impactor data are in terms of aerodynamic size.
In order to put the results on the same basis, it is necessary
to know the particle density so that one can convert physical
size to aerodynamic size (or vice versa).  In Figures 3 and 4
a value of 3 g/cm3 for density.has been used to convert the
physical diameter based on calculated diffusion battery penetra-
tions to penetrations based on aerodynamic particle diameter.
     The penetration plots Indicate an increase in efficiency
for particles smaller than 0.2 ymA. These results are consistent
with published literature in that  smaller, more highly diffused
particles experience increased collection efficiency.
     An overall penetration summary for runs 1 through 17 is pre-
sented in Table 4.  Total inlet and outlet mass loadings were
taken by cascade impactors.
                                 11

-------
    1.0
    0.5
    0.1
    0.05
I
W
    0.01
   0.005
   0.001 	
        0.1     0.2       0.5      1.0     2.0      5.0

              PARTICLE AERODYNAMIC DIAMETER, pmA

   Figure  3.   Penetration versus aerodynamic particle
              diameter for ductile operation.
                                                                      o.ooi
     0.1    0.2       0.5     1.0      2.0        5.0

           PARTICLE AERODYNAMIC DIAMETER,  umA
Figure 4.  Penetration versus  aerodynamic particle
           diameter for gray operation.

-------
             TABLE  4.   OVERALL PENETRATION SUMMARY
RUN NO.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
16P
17
OPERATION
(Gray)
(Ductile)
D
D
D
D
D
G
G
G
G
G
G
G
G
G
D
D
D
D
MASS LOADING *
mg/DNm3
INLET
	
2,050
2,100
1,430
1,810
2,060
1,910
2,230
2,310
1,880
2,100
2,290
2,260
2,320
1,960
1,790
2,250
1,390
OUTLET
23.6
20.9
24.5
28.0
27.0
32.0
30.7
35.8
58.5
39.3
52.9
27.9
24.0
23.0
21.2
24.8
24.8
21.2
OVERALL, PT
%
	
1.02
1.17
1.96
1.49
1.55
1.61
1.61
2.53
2.09
2.51
1.22
1.06
0.99
1.08
1.39
1.10
1.52
*N = 0°C, 1 atm
                             13

-------
     Although Figures 3 and 4 show penetrations for ductile and
 gray iron  operations separately, there was no significant dif-
 ference  in the penetrations between the two operations.  Likewise,
 the average  overall penetration for both the ductile operations
 and the  gray operation was 1.3%.
 OPACITY
     Opacity for the outlet stack of the National Dust Collector
 Model  850  variable rod module venturi scrubber was taken by em-
 ployees  trained and certified by the California State Air Quality
 Control  Section.  Readings were made five times a day from 8:00 a.m.
 to 4:00  p.m. every two hours.
     A summary of the daily average reading is presented below
 in Table 5.
                   TABLE 5.  OPACITY SUMMARY
                     Runs
                     2,3
                     4,5
                    6,7,8
                   9,10,11
                   12,13,14
                   15,16,17
                   Average
                   for Test
Average Opacity,
        13
        11
         8
        11
        25
        33
         2
        13
        15
ECONOMICS
     Cost data for the initial and annual cost was supplied by
the user.  The unit was originally built in 1967 as a fixed
hole orifice venturi and in early 1975 was modified to a variable
rod module type to alleviate plugging problems.  Shown on
Table  6  are the cost factors for the unit.  F.O.B. shipping
and delivery charges were not available, but should be about
$15,000.
                                14

-------
                    TABLE  6 .  COST DATA

Scrubber § Auxiliaries
    A. Fans, motors and motor starters            $136,200
    B. Ducting                                      88,000
    C. Liquid and solid handling and treatment      45,000
    D. Instrumentation                               7,800
    E. Electrical material                          55,000
                                           Total  $332,000

Scrubber Installation Cost
    A. Erection                                   $120,600
    B. Plumbing                                     61,900
    C. Electrical                                   36,000
    D. Engineering                                  15,500
                                           Total  $234,000
Total Initial Cost  (1967  prices)                 $566,000

Annual Costs include the following factors:
Operating Costs
    A. Utilities                                  $ 37,000
    B. Labor                                        20,000
    C. Supplies and materials                       12,000
    D. Treatment and disposal                        4,000
                                           Total  $ 73,000

Maintenance Costs
    A. Labor                                      $ 10,000
    B. Materials                                     6,000
                                           Total  $ 16,000

Plant overhead, space, heat,
  light, insurance                                $ _7,000
Total Annual Costs                                $ 96,000
                                15

-------
                         SECTION 4
            AMERICAN AIR FILTER KINPACTOR VENTURI

 SOURCE AND CONTROL SYSTEM
      The  emission source was a large borax fusing furnace used  in
 continuous operation.  The furnace was capable of producing  270
 metric tons per day of anhydrous borax CNa281,07) from the penta-
 hydrated  form of the feed, but was not always operating at full
 capacity  during the testing.  The particulates emitted were  pri-
 marily the hydrated and anhydrous forms of borax which escaped
 during the drying and fusing processes.
      The  total scrubbing system is shown in Figure 5, with the
 location  of the sampling ports indicated.  The gases from the fur-
 nace  at a temperature of 1,000°C to 1,100°C are quenched by  scrub-
 bing  liquor to about 80°C.  The gases then enter the American Air
 Filter Kinpactor 10 x 56 venturi scrubber.  The venturi is rec-
 tangular  in cross section with a throat height of 142 cm (56 in-
 ches) .  The throat width is automatically controlled to maintain
 a pressure drop of about 110 cm W.C. (43 in. W.C.) across the
 venturi.  The throat width can vary from 0.64 cm (0.25 in.)  to
 25.4  cm (10 in.).  Following the venturi the gases enter a cyclone
 entrainment separator.  The gas is moved through the scrubbing
 system by a blower which is rated at 44 Am3/s at 84°C and 114
 cm W.C. pressure differential (93,500 ACFM, 184°F, 45 in. W.C.).
 The blower is powered by a 746 kW (1,000 HP) motor.  The gases
 exhaust to the atmosphere through a 21 m tall, 2.13 m diameter
 stack.
     The scrubbing liquor is recycled through a tank which is
 fed about 28 l/s of fresh liquor or less err—amounts of fresh
water.  About 22 H/s of concentrated liquor is pumped from the
bottom of the tank so that the borax concentration of the liquor
 is maintained at from 10 to 15%.
                                 16

-------
                                                                             TO
                                                                        ATMOSPHERE
FROM
FURNACE
 	>
1,000°C
      LIQUOR
      41 i/s
      (MAX.)
             80°C
          20 m3/s
                              TO  LIQUOR
                         RECYCLE  41 fc/s
                            TANK  (MAX.)
   TO
RECYCLE
  TANK
BLOWER
                                                                                        STACK
                                                                                        70 C
                                                                                        25 m3/s
                 Figure 5,  Schematic diagram of scrubbing system.

-------
PROCESS CONDITIONS
     The gas conditions at the inlet and outlet sampling locations
for each run are presented in Tables 7 and 8.   The conditions
during runs 9 through 13 were thought to be most representative
of normal conditions because of the amount and consistency of
the product from the fusing furnace during these runs.  The
average barometric pressure during the testing period was 93.63
kN/m2  (27.65 in. Hg) .
     Close examination of the inlet and outlet flow rates indi-
cates a discrepancy in the measured flow rates.  The outlet flow
rate is greater than the inlet flow rate while the inlet temper-
ature is greater than the outlet temperature.   The static pressures
are practically equal.  There are two explanations for this flow
rate discrepancy: (1) As noted previously, the inlet flow is not
well developed and thus the inlet flow rate may not be accurate
and, (2) air could be leaking into the system  because of the low
pressure (-110 cm W.C. gage) within the system between the verituri
and the blower.
     The results on the Orsat analysis of the outlet gas are pre-
sented in Table 9.
CASCADE IMPACTOR DATA
     Particle size distribution data were obtained for the Ameri-
can Air Filter Kinpactor venturi scrubber as will be described in
the test method section.  Identical single point sampling at the
average velocity location was performed at both the inlet and out-
let.  The sampling time of each run depended on the mass loading.
The average sampling times for the inlet and outlet were nine and
fifty minutes respectively.
     Because of the number of very large particles in the inlet
gas and the entrained water droplets at the inlet and outlet, pre-
cutters were used.  The aerodynamic cut diameters of the inlet and
outlet  precutters were approximately 12 ymA and 4.5 ymA respective-
ly.  The inlet sampling was approximately isokinetic.  However,
because of the large amount of water droplets  in the outlet gas,
a special "rain can" had to be used on the outlet sampling nozzle.
                                 18

-------
            TABLE 7 .   INLET PROCESS CONDITIONS
Run
1
2
3,4
6,7
9,10
12,


,5
,8
,11
13
Temp. Water Volume
°C Percent*
75
108
73
78
79
81
36
57
14
23
49
41
Static Press. Flow Rate
cm W.C. ArnVs (ACFM)
-8
-6
-6
-7
-7
-8
.3
.5
.1
.2
.5
.6
24
21
16
18
20
20
(51
(45
(35
(38
(42
(43
,000)
,000)
,000)
,000)
,000)
,000)
            TABLE  8 .   OUTLET PROCESS CONDITIONS
Run
1
2
3,4
6,7
9,10
12,


,5
,8
,11
13
Temp. Water Volume
°C Percent*
80
73
72
68
71
69
31
33
38
28
23
21
Static Press. Flow Rate
cm W.C. Am3/s (ACFM)
-0
-0
-0
-0
-0
-0
.5
.4
.4
.4
.4
.4
29
25
25
25
25
25
(62
(54
(54
(54
(53
(53
,500)
,000)
,000)
,000)
,000)
,000)
*Based on wet and dry bulb temperatures
                            19

-------
        TABLE  9.  OUTLET AVERAGE GAS COMPOSITION
 Gas Component


      N2

      02

     C02

     CO

    H20

Molecular Wt.
Volume Percent
    (Wet)

     60

      6

      5

      0

     29

     26
Volume Percent
    (Dry)

     85

      8

      7

      0
     29
                        20

-------
Also, the outlet sampling nozzle was oriented perpendicular to
the gas flow.  The velocity through the outlet nozzle was main-
tained at the velocity of the outlet flow at that location.
     To minimize the possibility of condensation in the impactors
and to collect only the dry particles, the impactors were main-
tained at about 15°C above the gas stream temperature by using
heating blankets.  The precutters were not heated.
     The fact that the impactors were heated should be noted when
interpreting the size distribution data.  Some unpublished data
taken with heated and unheated cascade impactors in a stream of
wet borax particles have shown differences in the size distribu-
tions.  The primary difference was that the mass median aerody-
namic diameter of the borax particles collected in heated impac-
tors was as small as only 70% of the mass median aerodynamic
diameter of the particles collected in unheated impactors.  Thus,
the venturi scrubber may be encountering larger aerodynamic size
particles than the heated cascade impactor data indicate.
     Data from run 3 are not included because the outlet cascade
impactor stages had become wet.  This happened because an attempt
was made to sample with an outlet nozzle orientation parallel to
the gas  stream for this run.
Size Distributions
     The inlet and outlet size distributions both seem to be bi-
modal.   Thus, no attempt was made to fit all the data points to
a  log-normal curve.  The size distributions were approximately
log-normal  in the region below 2 ymA.  The mass mean geometric
diameters and geometric standard deviations for the log-normal
parameters  as well as the mass median diameter (from the data
points)  runs are presented in Table 10.
Diffusion Battery Data
     Diffusion battery data were taken during the fifth and sixth
days of testing.  The inlet data were taken during cascade impac-
tor runs 9, 10 and 11, while the outlet data were taken during
runs 12 and 13.

                                21

-------
      TABLE 10. PARTICLE SIZE DISTRIBUTION SUMMARY
Run


1
2
4
5
6
7
8
9
10
11
12
13

pm
ymA
0.80
0.94

0.86

0.92
0.82
0.66
0.86
0.76
0.94
0.75
Inlet
dpg
ymA
0.84
0.99
2.3
0.77
1.9
0.85
0.79
0.55
0.83
0.69
1.2
0.72

ag

2.1
3.3
3.7
2.7
3.3
4.7
3.3
2.7
3.3
2.7
3.3
3.0

pm
ymA
0.40
1.1
0.32

0.33

0.23

--
0.25


Outlet
Sg
ymA
0.55
0.78
0.32
0.22
0.41
0.20
0.22
0.13
0.16
0.25
0.16
0.19

ag

2.0
3.2
2.6
2.9
2.1
2.8
3.4
3.6
3.5
5.2
3.6
3.5
Note:  d   = mass median aerodynamic particle diameter
        p    from data

       d   = log-normal geometric mass mean aerodynamic
        P8   particle diameter

        0  = log-normal geometric standard deviation
                          22

-------
     Because of the large amount of water vapor  (40% by volume)
in the gas, the lenses in the condensation nuclei counter of the
diffusion battery would fog when the diffusion battery was opera-
ted in the normal manner.  Increased dilution did not solve the
problem because the particle count would then drop below the
threshold of the counter.  Also, heating the diffusion battery
to its maximum allowable temperature in an effort to reduce con-
densation on the lenses did not help.
     The system was modified to allow data to be taken by routing
the incoming source gas through a glass flask before entering the
diffusion battery.  Enough of the water vapor condensed in the
flask so that the condensation nuclei counter did not become in-
operable.  Any condensation of water vapor would cause collection
of submicron particles by diffusiophoresis.  Thus, a fraction of
the particles did not reach the diffusion battery.  For this rea-
son the diffusion battery data may not be accurate.
     However, since the configurations for taking the outlet data
and the inlet data were the same and the process was constant
during the  testing period, the data still give an indication of
the relationship between the distribution of submicron particles
at the inlet and outlet.  The data show that the outlet submicron
particles  are larger  and more monodisperse.  This is an indication
that particle growth  by condensation may be occurring in the scrub-
ber .
      The  data are only qualitative because condensation also oc-
curred within the measuring system.  However, mechanisms for par-
ticle growth are present in the scrubber system.  These mechanisms
are discussed later.
PARTICLE  PENETRATIONS
      Particle penetration versus particle aerodynamic diameter
was calculated  from the cascade impactor data.   The results are
shown in  Figures 6 through 10 for each day of testing.  Pene-
trations  for run 1 are not shown because they are much larger
than  the  penetrations for all the other runs, indicating anomalous

                                 23

-------
ts)
      O
      I—I
      E-i
      U
O
I—I
H


W

W
           0.05
           0.01
          0.005
          0.001
               0.3   0.5
        PARTICLE AERODYNAMIC DIAMETER,  d  ,ymA
                                         pa

       Figure  6. Particle  penetration for run 2,
                                                 O
                                                 i—i
                                                 H
O
i—i
H
                                                       w
                                                           0.05
                                                       SS   o.oi
                                                          0.005
                                                    0.001
                                                         0.3  0.5     1        23

                                                   PARTICLE AERODYNAMIC DIAMETER,d  ,
                                                                                 '  pa'
                                                  Figure 7. Particle penetration for runs
                                                            4 and  5.

-------
ts)
tn
o
I—I
H


i
PL,

 A



I—1
H


H


W
            0.05
            0.01
           0.005
           0.001
                0.3  0.5     1      23


           PARTICLE AERODYNAMIC DIAMETER,d  ,ymA
                                          pa

     Figure  8.  Particle penetration for runs

                6,  7,  and 8.
    0.05
                                                     O

                                                     H


                                                     §
O
I—i

H




W
    0.01
   o.oos £EE:
                                                        o.ooi
        0.3  0.5     1        23


  PARTICLE AERODYNAMIC DIAMETER,d    umA
                                 pa*

Figure 9. Particle penetration for  runs
          9, 10 and 11.

-------
  2
  O
CJ


2
p.,

 «*


O
I—I
H

2
H



w
        0.1
       0.05
       0.01
     0.005
     0.001


          0.3  0.5     1        23


     PARTICLE AERODYNAMIC DIAMETER, d
                                     pa
Figure 10. Particle penetrations  for  runs  12  and 13.
                     26

-------
behavior.  Penetrations for runs from the last two days, when
operations were most smooth, are quite consistent.
     Because the size distributions were not log-normal all of
the penetrations were calculated manually.  Particle penetration
based on diffusion battery data is not presented because of the
inaccuracies incurred during the data acquisition.  Diffusion
battery data are discussed in the previous section.
     The total mass loadings and overall penetrations for the
runs are presented in Table 11.  The total mass loading was deter-
mined from analysis of the cascade impactor data.  Run 1 had ano-
malously high overall penetration and high flow rates which indi-
cated that something may have been wrong with the data or that
the venturi was not operating properly during that run.  The
average penetration for all of the runs, exclusive of run 1, was
2.5%.

           TABLE  11.  MASS LOADING AND OVERALL PENETRATION
Run
1
2
4
5
6
7
8
9
10
11
12
13
Inlet Mass Loading
mg/DNm3
332
653
1,150
841
690
2,310
882
829
1,040
863
1,210
1,000
Outlet Mass Loading
mg/DNm3
40.4
18.3
42.1
11.0
21.9
18.4
20.5
22.7
19.3
31.0
36.9
27.3
Penetration
1
12.2
2.8
3.7
1.3
3.2
0.8
2.3
2.7
1.9
3.6
3.0
2.7
                                 27

-------
OPACITY
     Opacity for the outlet stack of the American7Air Filter
Kinpactor 10 x 56 venturi scrubber was  determined by personnel
trained and certified by the California Air Resources Board.
Readings were made hourly during  the testing period.   The opa-
city determinations were made somewhat  difficult by the presence
of steam condensation in the plume and  the  proximity of other
stacks emitting similar particulates.
     Table 12 presents the daily average opacity readings.
                      TABLE  12.   OPACITY
   Date
Runs
Average Opacity,
Average Outlet
Loading, mg/m
8/19/76
8/20/76
8/21/76
8/22/76
8/23/76
2
3-5
6-8
9-11
12,13
10-15
15
20
15-20
20
9
12
11
14
19
                                28

-------
ECONOMICS

     Data for the initial costs of the venturi-cyclone scrubbing
system, purchased in 1970, were supplied by the user:

     Approximate scrubber purchase cost             $200,000
     Scrubber auxiliaries:
        1. Fans, motors, etc.                         30,000
        2. Ducting                                    47,000
        3. Liquid and solid handling and
           treatment                                  50,000
        4. Instrumentation                            45,000
        5. Electrical material                        36,000
     Scrubber installation:
        1. Site preparation                          108,000
        2. Installation                              100,000
        3. Startup and modification                   51,000
        4. Engineering                                63,000
                            Total Initial Cost  -   $730,000
     The operating costs were not available.  However, the power
 costs  can be estimated.  The major power user is the large blower,
 The blower  is powered by a 746 kW (1,000 HP) motor.  At $0.03
 per kW-hour the fan power would cost $22 per hour, or $537 per
 day.
                                 29

-------
OPERATING PROBLEMS
     The primary operating problems with the system are the plug-
ging of nozzles -et&d piping and scale build-up in the system.
These problems are all caused by calcium carbonate and sulfate
deposits.  The local water is very hard and the feed to the fusing
furnaces may also contain these mineral impurities.  To combat
this problem special reamer nozzles have to be used and the system
has to be shut down periodically to chip away the built-up scale.
     Although a large amount of entrained water was carried-over
from the entrainment separator through the fan, and out the stack,
it was not considered a problem by operating personnel.
                               30

-------
                            SECTION  5
                   GAS-ATOMIZED SPRAY SCRUBBER

SOURCE AND CONTROL SYSTEM
     The emission source for this test was a no. 7 gray iron
cupola used for an 8 to 9 hour/day  operation.  The cupola nor-
mally is run for 3 days/week, but during the test period it
was run for as many days as possible  (3 days in the first week,
5 days in the second week).  The operation produces 10 to 12
tons per hour of metal from a feed  of scrap metal, coke, and
limestone.
      During the testing period the cupola ran at normal condi-
tions.  The particulate emissions result from the melting of the
scrap and burning of the coke in the cupola.  Figure 11 is a
schematic diagram of the entire operation including the scrubber
system.
     The hot gases from the melting operation enter the after-
burner where it reaches approximately 700 - 1,000°C.  The gas
is  then quenched by city water, which during the test had a
flow rate varying between 0 and 2.6 £/s (4.1 GPM) giving a
temperature range of 700°C  to 60°C.  Scrubber liquor is added
just ahead of the throat section (the rate was varied during the
testing period from 2.5 £/s to 15.8 £/s).  Next the gas enters
a variable throat scrubber  section which is a rectangularly cross-
sectioned orifice with a throat height of 91.4 cm (36 inches).
The throat width  can be varied from 4.2 cm (1.7 in.) to 11.4 cm
(4.5 in.).  Following the venturi the gas enters a cyclone type
entrainment separator.  An  induced draft blower moves the gas
through the scrubbing system, powered by a 298 kW (400 HP) motor.
From this blower the gas is exhausted through an 11.9 m (39 ft)
high, 0.91 m (3 ft) diameter stack.
     At the venturi section, scrubber liquor is pumped in through
pipes placed directly upstream of the throat.   This liquor

                                31

-------
            WET CAP
   CHARGING
      DOOR
CM
to
                                           QUENCH
                                           SPRAYS
                    SCRUBBING
                     LIQUID
                                   C
                  ENTRAINMENT
                   SEPARATOR
             CUPOLA
AFTERBURNER QUENCHER
                                                                   SCRUBBER  SUMP
                                                                                                 t
STACK
                           Figure  11.   Schematic  drawing  of  scrubber  system.

-------
begins as clean water every day from a  sump  below  the  cyclone.
No fresh water feed is added to this liquor  so  that  as  the  day
progresses the solids concentration increases.  When the  system
is shut down the sump is neutralized, cleaned out, and  fresh
water pumped in for the next day of operation.
     The water used for the quench sprays, however,  comes directly
from the city water lines and no recirculation  is  used.
PROCESS CONDITIONS
     The scrubber process conditions for  the inlet and  the  outlet
are shown in Tables 13 and  14  .  Runs 1 through 4  represent the
scrubber running with a number of leaks,  as  can be seen by  com-
paring the inlet and outlet flow rates.   Between runs 4 and 5
many of the large leaks were patched, resulting in much smoother
operation as well as a reduction of the fan  amperage, an  increase
in the inlet flow rate, and reduction in  the outlet  flow  rate.
     After patching most of the leaks,  more  control  over  the
scrubber operation could be maintained.   It was then possible
to adjust water flow rates  as well as the pressure drop across
the venturi.  This enabled  a number of  test  conditions  to be
examined.  For  the most part, the charging operation was  fairly
constant and did not vary significantly  between  the runs,  except
for run  1 when  the crane broke and only one  charge was  added
during the entire run.  The average barometric  pressure during
the testing period was 99.36 kPa  (29.34 in.  Hg).
      The  top of the cupola was open, except  for a  "wet  cap."
The wet  cap consists of baffles and sprays at the  top of  the
cupola and it was operated  during the test period.   Depending
upon  the  scrubber conditions, varying quantities of  visible
emission were observed flowing from the wet  cap.
      The  tests  were conducted during the  winter so that the am-
bient temperature averaged  about 0°C during  the testing.  This
low ambient temperature probably caused the  temperature of  the
gases entering  the scrubber to be lower than normal  because of
heat  transfer through duct walls, leakage of cold  air,  and  use
of cooler than  normal quencher and scrubber  water.

                                33

-------
                 TABLE 17,. SCRUBBER CONDITIONS
                                                                                         TABLE 14. PROCESS CONDITIONS
Run
No.
1
2
3
4
5
6
1-B
7
8
2-B
9
10
11
12
13
14
15
Pi
*
-1.
-1.
-3.
-2.
-6.
-6.
-6.
-6.
-3.
-3.
-5.
-4.
-2.
-2.
-2.
-5.
-5.
8
3
0
0
4
4
6
1
8
6
1
3
3
5
8
8
8
P2
*
- 77
- 81
- 57
- 64
- 83
- 85
- 87
- 87
- 86
- 85
- 87
- 85
-105
-105
-107
- 94
- 92
AP
*
.5 75
.3 80
.2 54
.8 62
.8 76
.1 78
.6 81
.6 81
.4 82
.1 81
.6 82
.1 80
.0 103
.0 104
.0 104
.0 88
.7 86
.7
.0
.2
.8
.4
.7
.0
.5
.6
.5
.5
.8
.0
.0
.0
.2
.9
Throat
Width
(cm)
11
4
5
4
11
11
11
11
S
5
5
5
4
4
4
11
11
.4
.4
.6
.4
.4
.4
.4
.4
.1
.1
.6
.6
.2
.3
.3
.4
.4
Scrubber
Liq.Flow
(Vs)
3
11
< 2
< 2
12
12
12
12
9
9
9
9
12
12
12
15
15
.2
.4
.5
.5
.6
.6
.6
.6
.5
.5
.5
.5
.6
.6
.6
.8
.8
Quench
Flow
O/s)
1
0
1
1
1
1
1
1
1
1
0
0
2
2
2
1
1
.1
.0
.6
.6
.6
.6
.6
.6
.6
.6
.0
.0
.6
.5
.5
.7
.7
Pj = static pressure before venturi
P2 = static pressure after venturi
AP = Pi- P2
 * = numbers expressed in cm W.C.
Run
No.
1
2
3
4
5
6
1-B
7
8
2-B
9
10
11
12
13
14
15
Inlet
Temp.
°C
246
496
132
132
177
149
149
177
177
177
699
677
68
60
71
66
177
Outlet
Temp.
°C
56
56
57
57
65
65
65
54
54
54
58
58
51
51
51
51
51
Inlet Outlet Inlet
Water Water Flow Rate
Vapor Vapor Am /s
Vol.% Vol.% (MACFM)
**
21
11(6*)
27
27
25
26
26
25
25
25
2*
2*
27
20
32
25
25
7
7
17
17
20
20
20
5
S
5
S
5
8
8
8
11
11
5.2(11.0)
6.2(13.2)
6.5(13.7)
6.5(13.7)
11.4(24.1)
11.0(23.4)
11.0(23.4)
10.8(22.9)
10.8(22.9)
10.8(22.9)
12.6(26.7)
12.5(26.4)
4.1( 8.7)
4.0( 8.4)
4.2( 8.8)
9.1(19.2)
10.4(22.2)
Outlet
Flow Rate
Am /s
(MACFM)
** A
11.7(24.8)
11.7(24.8)
12.5(26.5)
12.5(26.5)
10.5(22.2)
10.5(22.2)
10.5(22.2)
10.0(21.1)
10.0(21.1)
10.0(21.1)
8.1(17.1)
8.1(17.1)
7.0(14.8)
7.0(14.8)
7.0(14.8)
9.8(20.7)
9.8(20.7)
  * Based on sampling train water catch
 ** Based on wet and dry bulb temperatures
*** MACFM = thousand actual cu.  ft/min

-------
     The cyclone separator was not operating  efficiently  during
the testing period.  It was patched along with the rest of the
system after run 4 but it could not be sealed completely  as it
had been repeatedly patched previously.  Any  leakage into the
cyclone would disrupt the flow of liquid down the walls and
would cause reentrainment.  Entrainment emission was clearly
noticeable as rainout from the stack.  The entrainment emission
was due to several factors such as leakage, poor flow distribu-
tion at the inlet, and internal roughness due to scale buildup.
     The Orsat analyses of the outlet gases are presented in
Table 15.
CASCADE IMPACTOR DATA
     Particle size distribution data were obtained for the gas-
atomized spray scrubber as described in the test method section.
Identical single point sampling at the average velocity location
was performed at both the inlet and outlet.  The sampling time
for each run depended upon the mass loading.  The average sampling
times for the inlet and outlet, respectively, were 6.5 and 31
minutes.
     The inlet sample ports, because of a limited length of
ducting after quencher and before the venturi throat, were loca-
ted 0.73 equivalent duct  diameters downstream of a 78° bend and
0.73 equivalent duct diameters from the beginning of the venturi
section.  The inlet duct  was rectangular  (0.91 m x 0.61 m) and
 its equivalent duct diameter  was 0.73 m as calculated by EPA
Method  1.
     The outlet sample ports were located in a 0.91 m (3 ft)
diameter round stack, about 8 stack diameters downstream of the
inlet from the fan and 3  diameters upstream of the stack air.
The velocity traverses indicated a well-developed flow pattern.
     Entrained water drops were a problem at both the inlet and
outlet, so precutters for the inlet and outlet were approximately
11.6 ymA and 5.6 ymA, respectively.  Both the inlet and outlet
sampling were approximately isokinetic.
     To minimize the possibility of condensation in the outlet

                               35

-------
TABLE 15. AVERAGE GAS COMPOSITION
Gas
Component
N2
02
C02
CO
Molecular Wt.
Inlet
Volume,!
Dry
80.1
6.4
13.2
0.3
29.9
Outlet
Volume, 1
Dry
(Runs 1-4)
80.1
16.9
3.0
0
28.8
Outlet
Volume,!
Dry
(Runs 5-15)
79.6
14.6
5.8
0
29.2
                 36

-------
impactors and to collect dry particles,  the  impactors  (except
run 9 which was below the  stack  temperature  due  to  a malfunction
of the heating blanket) -were maintained  at about 13.5°C  above  the
gas stream temperature using heating blankets.   Because of the
temperature fluctuation at the inlet the impactor was kept out of
the stack and heated by a heating-blanket to an  average tempera-
ture of 86°C.  In both cases (inlet and outlet)  the precutters
were not heated.
     The fact that the impactors were heated should be remembered
when interpreting the size distribution data.    Because of the
presence of  a quencher upstream  of  the scrubber  section, the
particles may be wet due to condensed or absorbed water when
they enter the  scrubber.   The wet particles  would have a different
size distribution  than those collected in a  heated  impactor.  The
wet particle size  distribution would probably  have  a larger
mass median  diameter.   The runs  with the most  favorable  con-
 ditions for  wet particles  were  runs 11-14.
     Table  16 summarizes the "d   "  and "a "  results for the
                               to        o
cascade impactor runs using log-probability  parameters.  The
distribution data  include  only the  results of  the cascade impac-
tor stage  analysis.  The precutter  and probe weight gains were
not used.  The  log-normality approximation was adequate only for
diameters  between  0.2 and  2  ymA.
DIFFUSION  BATTERY  DATA
      Diffusion battery data were taken during  cascade impactor
 runs as shown in Table  17.
     A dilution system was used  for this testing along with the
A.P.T. diffusion battery.  A 208 liter (55 gallon)  drum was used
upstream of  the diffusion  battery as the dilution system.
Samples were not taken continuously from the stack because
the intermittent charging  operation can  result in rapidly chan-
ging readings.  A  syringe  was inserted into  the  stack several
times  to withdraw  gas samples and these  were expelled into  the
dilution tank.  Later this system was switched to a calibrated
hand pump.   These  gas samples were  diluted with  dry filtered  air

                               37

-------
                  TABLE 16.   CASCADE IMPACTOR DATA USING
                            LOG-PROBABILITY ANALYSIS
Run No .
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Inlet
dpg
(ymA)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
.47
.45
.40
.26
.19
.33
.39
.45
.58
.37
.40
.41
.31
.20
.18
Inlet
°g
2
2
2
2
3
2
2
2
2
2
2
2
2
4
4
.6
.9
.1
.7
.4
.0
.0
.0
.0
.7
.4
.4
.7
.4
.4
Outlet
dpg
(umA)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
.86
.76
.38
.58
.36
.43
.41
.39
.49
.35
.53
.43
.41
.43
.49
Outlet
°g
1
4
2
2
2
2
2
2
2
2
2
2
2
2
2
.9
.0
.3
.3
.3
.2
.2
.5
.3
.7
.0
.7
.4
.4
.2
oo
TABLE 17.   DIFFUSION BATTERY PARTICLE
           SIZE DISTRIBUTIONS
i 	
D.B.
Run
No.
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Impactor
Run No.
5
5
6
1-B
8
8
2-B
2-B*
2-B*
2-B*
11
11
12
13
Type
of
Run
In
In
In
In
In
In
In
Out
Out
Out
Out
Out
In
In
V
Cum)
0.070
0.092
0.095
0.128
0.072
0.095
0.094
0.057
0.061
0.085
0.090
0.090
0.086
0.059
a
g
3.4
12.0
12.0
14. '0
6.0
2.9
11.5
1.0
2.5
4.5
6.0
8.9
9.3
6.0
Total
Particle
Count**
No. /cm3
9.0 x 106
1.1 x 107
1.1 x 107
3.9 x 106
2.3 x 106
1.3 x 10s
1.1 x 107
5.1 x 106
8.0 x 106
8.3 x 105
7.8 x 105
1.9 x 106
2.2 x 107
1.5 x 107
                                                                        *Same condition but not simultaneous

                                                                       **Gas conditions: 20°C, 1 atm

-------
in the tank and run through the diffusion battery with particle
counts being measured using the condensation nuclei counter.
     The particle size distributions determined by diffusion
battery analysis are given in Table 17.
PARTICLE PENETRATIONS
     Particle penetrations for various particle aerodynamic dia-
meters were calculated from cascade impactor and diffusion bat-
tery cumulative loading data.  The penetrations based on diffusion
battery data were calculated from log-normal inlet and outlet
cumulative distributions.  The physical (actual) size distribu-
tions from the diffusion battery analysis were converted to
aerodynamic size by assuming the cupola dust had a particle
density of 2.5 g/cm3.
     Since the size distributions were not log-normal, the
cascade impactor penetrations presented in Figures 12 through
18 were calculated graphically.
     The total mass loadings and overall penetration for the
runs are presented in Table 18.  Mass loading and penetration
are also shown without the outlet precutter catch.  The reason
is that the precutter loading was mostly entrainment carryover
from the cyclone and was not due to scrubber inefficiency.
                                39

-------
                  TABLE  18.  MASS LOADING AND
                             OVERALL PENETRATION
Run
No.
1
2
3
4
5
6
1-B
7
8
2-B
9
10
11
12
13
14
15
Mass
Inlet
1,570
2,120
1,100
2,060
3,190
2 , 340
7,030
2,530
1,913
1,600
5,070
2,620
1,750
2,370
2,500
1,450
1,240
Loading, mg/DNm *
Outlet i
212
74.3
280
176
609
341
968
234
320
177
198
589
159
214
797
278
277
Outletz
46.3
36.5
267
175
514
307
-
234
98.2
-
156
159
153
159
174
186
223
Penetration, \
Pt1
13.5
3.5
25.5
8.5
19.1
14.6
13.8
9.3
16.7
11.1
3.9
22.5
9.1
9.0
31.8
19.2
22.2
Pt2
2.9
1.7
24.4
8.5
16.1
13.1
-
9.2
5.1
-
3.1
6.1
8.7
6.7
6.9
12.8
17.9
  Including outlet precutter dry weight gain

  Without outlet precutter dry weight gain

* N = 0°C, 1 atm
                             40

-------
  1.0
0.01
                                              10
      PARTICLE AERODYNAMIC DIAMETER, d
                                        .
                                      pa
§
1-1
H


%
                                                                 1.0
                                                                  0.5
                                                                  0.2
                                                                  0.1
                                                             H
                                                             w

                                                             §  °-05
                                                             Cu
                                                                0.02
                                                                0.01
Figure 12.  Particle penetration for cascade impactor

            runs 1 and 2.
        0.2      0.5     1       2         5       10


           PARTICLE AERODYNAMIC DIAMETER, d   . ymA
                                           pa


   Figure 13. Particle penetration for cascade impactor

              runs 3 and 4.

-------
 1.0
                                                                          1.0
0.03
    0.05   0.1
 0.2

PARTICLE AERODYNAMIC DIAMETER, d
                                                  lpa'
Figure 14.  Particle penetration for cascade impactor runs
            5, 6 and 7 and diffusion battery runs 3-9 (inlet)
            and 10-12 (outlet).
                                                                         0.01
    O.OS   041     0.2       0.5     1      2         5

           PARTICLE AERODYNAMIC DIAMETER, d  , pmA
                                           pa

Figure 15.  Particle penetration for cascade impactor run 8.
                                                                                                                                      10

-------
                                                                        1.0
CM
         1.0
       0.01
           0.2
              0.5     1       2         5


        PARTICLE AERODYNAMIC DIAMETER, d  .
                                        pa


Figure 16.  Particle penetration for cascade impac-tor

            runs 9 and 10.
                                                                      0.01
                                                                      .005
                                                                          .05     0.1    0.2       0.5     1       2         5      10


                                                                                    PARTICLE AERODYNAMIC DIAMETER, d  , ymA
                                                                                                                    pa


                                                                       Figure 17.  Particle penetration for cascade impactor runs 11,

                                                                                   12 and 13, and diffusion battery runs 15 and 16

                                                                                   (inlet) and 13 and 14 (outlet).

-------
0.01
                                               10
        PARTICLE AERODYNAMIC DIAMETER, d  , umA
                                     '  pa' ^

Figure 18.   Particle penetration for cascade impactor

            runs 14 and 15.
                   44

-------
OPERATING PROBLEMS
     The primary operating problem with the system at first was
the limiting capacity of the fan.  Great amounts of excess air
were getting into the system through all of the leaks, resulting
in the fan running at its maximum capacity.  For this reason the
system could not pull enough gas through the scrubber and a large
portion left through the top of the cupola.  The amount of water
that could be added to the system was also limited, which in turn
affected the scrubber efficiency.
     Once most of these leaks were patched, however, the fan
amperage dropped well below the limit, and the system could be
adjusted easily.
     The scrubber liquor was another operational problem.  Since
the cyclone entrainment separator was not working effectively, much
of the water which captured the particles was carried over, through
the fan and out the stack.  The scrubber liquor was recycled and
changed only at the end of each day which meant that the solids
concentration increased as the day progressed.  Reentrainment of
this water, therefore, would mean higher outlet loadings since
this extra weight was considered in the emission rates.
     Corrosion had a severe effect on the system operation because
it was  the cause of the excessive leakage of air into the scrub-
ber and  entrainment separator.
     Another problem with the system was the absence of a cap
on the  cupola top.  This allowed the gases to escape through the
 top.   Some of the gases do not pass through the scrubber  and
 therefore are not treated effectively, which in turn caused a
visible emission problem.  This was alleviated by capping the
cupola  top subsequent to our tests.

ECONOMICS
     The scrubber system was built about ten years ago with sub-
stantial participation by the foundry.  Consequently, the cost
                                 45

-------
was distributed among several purchased components, contracted
work, and foundry-performed work.   Records of the costs were
not kept in one place and it was not possible to obtain precise
data on the sub-system costs.
     The operating cost for the blower (300 kW motor]  was about
$72.00 per day, based on eight hours use per day.
                               46

-------
                           SECTION  6
                    PERFORMANCE  COMPARISON

INTRODUCTION
     All three scrubbers evaluated fall within the general classi-
fication of gas atomized spray scrubbers.  The AAF venturi is
most like a conventional venturi scrubber.  The performance of
gas atomized scrubbers depends on many factors which involve
both the properties of the dust  to be collected and the operating
conditions of the scrubber.  Yung, et al. (1977a) have modeled
the performance of Venturis assuming collection results from par-
ticle impact ion on the liquid drops atomized in the venturi throat.
The important parameters are particle aerodynamic impaction dia-
meter, liquid drop diameter, amount of liquid, and relative velo-
city between the particles and the liquid.  The energy used by a
venturi goes primarily into achievement of high relative velocities
and is manifested in the pressure  loss.
PRESSURE LOSS
     Yung, et al.  (1977b)  have derived a  relatively simple model
for the pressure loss  in a venturi:

               AP = 1  x 1(T3 u*e  u^ [-iij              (1)
where:  AP = pressure  drop,  cm W.C.
             ratio  of  the velocity  ;
             at the throat exit  to  the gas velocity, dimensionless
Ul  = ratio of the velocity attained by the liquid drops
 de
        Gt
    = gas velocity in throat, cm/s
 QT  = liquid volume flow rate, cm3/s
 Qr = gas volume flow rate, cm3/s
                                 47

-------
This model is applicable to large Venturis in which  the predomi-
nant energy loss mechanism is the acceleration of the  liquid
drops.  The term "u* " is not ususally measured and can be con-
sidered an empirical constant.  It has a direct mathematical
relation to the venturi throat length under certain conditions,
however.  If the drop acceleration takes place in a constant area
section and the initial drop velocity is insignificant then,
                 u*e = 2[1 - x2 + (x* - X2)0'5]           (2>
                 3£. C-  pr
where:   x = 1 +   t  Do  G                              (3)
                  16 dd PL

        i. = venturi throat length, cm
        C,,  = drop drag coefficient at throat inlet, dimensionless
        PG = gas density, g/cm3
        d, = drop diameter, cm
        p, = liquid density, g/cm3

For simplicity the drop diameter is taken as the Sauter mean
diameter resulting from gas atomization as predicted by the
Nukiyama-Tanasawa relation.
PARTICLE COLLECTION
     Yung, et al. (1977a) found that the assumption that particle
collection occurs primarily in the constant area throat section
is often valid and greatly simplifies the performance prediction.
In the throat section the particle penetration reduces to:
                                48

-------
  In Pt(d )
     	 =	.	   4 K    n   *  ^ I-5            *
     B        K  fl-ii* 1 + n i     PO  u~udej    +  4.2  (1-u,  )°'5
                           0.7
de-
                                                         (4)
 where:        B  =
           Kp0  ~	 = inertial parameter based  on
                       G  d      throat velocity,  dimensionless
                                                         (6)

 and,   Cr = Cunningham slip  factor,  dimensionless
       d  = particle  diameter,  cm
       p  = particle  density,  g/cm3
       Ug = gas viscosity, poise

     Equation (4) sometimes  slightly underestimates  the particle
 collection occurring  in a  venturi scrubber.  For most industrial
 venturi scrubbers, particle  collection can be predicted closely by
 neglecting the first  term  in the  right hand side of  equation  (4).
     The overall penetration prediction requires the integration
 of equation (4)  over  the inlet  dust  particle size  distribution.
VARIABLE ROD MODULE VENTURI  SCRUBBER COMPARISON
     The variable rod module venturi  is  essentially  several rec-
tangular jets  in parallel.
                                 49

-------
Pressure Drop
     The exact geometry of the venturi rod module was not made
available.  Thus, throat length and throat gas velocity were not
known, and the measured pressure drop could not be compared to
that predicted by equation (1).
Particle Collection
     Since "u*i " was not known from the geometry, penetration pre-
dictions for a range of values were made and are shown in Figures
19 and 20.  The conditions used are shown in Table 19. The closest
fit to the data occurred for Uj   =0.75 which corresponds to a
throat length of 27 cm and a throat gas velocity of 11,700 cm/s,
using equations (1) to (3).

           TABLE 19. CONDITIONS  FOR VARIABLE ROD VENTURI
                     PERFORMANCE PREDICTIONS
          AP = 178 cm W.C.
       QL/QG = 0.00168
          PL = 1 g/cm3
          pr = 0.9 kg/m3
yG = 1.89 x 10"
ude =
uGt '
dd '
CDo '
B =
*t =
1.0
10,140
107
0.56
3.32
oo
g/cm-s
0.75
11,710
101
0.55
3.42
27

0.5
14,340 cm/s
93 ym
0.53
3.52
6.3 cm
     Based on an approximate angle of free jet divergence of 20°
and a spacing between rods equal to the rod radius, it was es-
timated that the rod diameter was 7.2 cm (2.8 in.), which was
within reason.
     The overall penetration was calculated on averaged cascade
impactor data.  For gray iron operation (d   = 0.91 ymA, a  = 1.6)
                                          XT O              O
                                 50

-------
O
U
o
l-
w
 w
 IX
    0.1
    0.01
    0.001
         0.1
                  PARTICLE AERODYNAMIC  DIAMETER,  pmA
    Figure 19.   Comparison of predicted and measured penetrations
                for the NPC venturi during gray  iron operation.
                                                                             0.1
                                                                         o
W

W
   0.01
                                                                           0.001
                                                                                 0.1
                 PARTICLE AERODYNAMIC  DIAMETER,  ymA
  Figure  20.   Comparison  of predicted  and  measured penetration
              for  the NDC venturi  during ductile  iron operation

-------
a penetration of 4.1% was predicted and for ductile iron operation
(d   = 0.57 ymA, a  = 1.7) a penetration of 16.2% was predicted.
  fT ^?              v5
The data indicated the overall penetration was between 1% and  2.51,
which meant that the size distribution used in the predictions
was not correct for sizes larger than about 3 ymA.
Discussion
     The model assumes that collection is by inertial impaction
on drops, occurring only in the throat region of the venturi.  The
data agree with the model in the range of 0.8 to 2.0 umA diameters
but show greater efficiency below 0.8 ymA.   Thus, the model is
fairly close for the larger particles of interest.  However, as
particle diameter decreases below 1 ymA diameter, forces other
than inertia,  such as  flux forces  and Brownian motion,  become
important.
     Diffusiophoresis, and thermophoresis,  two of the flux force/
condensation (F/C) mechanisms are effective in regions where the
liquid spray drops are cooler than the gas  and the gas is satura-
ted.  Some condensation may have caused particle growth between
the inlet sampling port and the variable rod module.  Here the pre-
scrubber sprays may have been a few degrees cooler than the gas,
so that enough particle growth could have occurred prior to the
venturi rod module to increase the collection efficiency of the
smaller particles.
     F/C effects could have caused the cooling spray drops to
collect fine particles in the cooling section of the tower.  The
cooling spray water temperature was about 20°C lower than the gas
temperature.  Collection by Brownian diffusion would not be im-
portant in the size regime that was measured in these tests.
     Another section of the system where small particle collection
could have occurred was the section containing the two blowers
in series.  The high velocity and extreme turbulence in the blow-
ers may have been effective in causing particle collection.
AAF KINPACTOR VENTURI SCRUBBER COMPARISON
     The American Air Filter venturi is a classical venturi with
a rectangular throat.   The throat width is  automatically controlled

                                52

-------
to maintain a pressure drop of about  110  cm W.C.
Pressure Drop
     Since the throat width was  automatically  controlled the
throat area and gas velocity were not known with any precision.
The maximum throat area possible was  0.361 m2.  The throat length
was not known either.  Thus, direct comparison of pressure drop
measured with equation (1) was not possible.
Particle Collection
     In many Venturis the ratio  of drop velocity to gas velocity
at the throat exit (u*[e) is about 0.8, and this value will be
used in the prediction.  The liquid to gas ratio used is based
on the gas flow rate at the outlet for runs 12 and 13 (20 Am3/s)
and 80% of the stated liquid flow rate  (33 £/s) .  Table 20 presents
the conditions used for the performance prediction.

              TABLE  .20. CONDITIONS FOR AAF VENTURI
                        PERFORMANCE PREDICTION
                    AP  =  110  cm W.C.
                QL/QG  =  0.0017
                    PL  =  1.0  g/cm3
                    PG  =  0.79 kg/m3
                    yfi  =  1.6  x 10""  g/cm-s
                   ude  =  °'8
                   ur+  =  8,860 cm/s
                    bt
                    dd  =  104  urn
                   CDO"  Q'58
                     B  =  3.37
                    j,  =  41 cm
                     T-	

      The comparison of  prediction  with the data  is  shown  in
  Figure 2L  The measured penetration is also shown corrected  for
                                  53

-------
  2
  O
  t—i
  H
  U
2
O
i—i
H


H


W
       0.05
                    PREDICTION
                    CORRECTED FOR
                    WET PARTICLES
                    WITH DILUTION
              MEASURED  FOR
              DRY PARTICLES
       0.01
      0.005
      0.001
           0.3  0.5     1       23

    PARTICLE AERODYNAMIC DIAMETER, d__,
                                    pa

Figure 21.   Comparison of predicted and measured
            penetration for the average of runs 12
            and 13 of the AAF venturi
                    54

-------
 dilution and for a diameter   increase  to  account for the fact
 that the measurements were made with heated  impactors.  Based
 on an average dp? = 1 umA and ag =  3 for  the inlet dust the
 overall penetration was predicted to be about 15% while the
 measured penetration was about 3%.
Discussion
     In order to compare the model with experiment a few important
factors need to be noted:
     1. Heated impactors - As explained previously the inlet and
outlet size distributions were determined  using heated cascade
impactors.  Based on previous experience it was expected that the
mass median diameters of the actual  wet borax particle size dis-
tributions were up to 1.5 times the  sizes  measured in the heated
impactors.  The high solubility of borax was  the primary reason
for the difference.  Thus the  venturi was  collecting larger par-
ticles than those measured in  the heated impactor.  In order to
compare the experimental results with the  prediction it is neces-
sary to plot experimental penetration against the actual (wetj
particle size rather than the  dried  particle  size.  This correc-
tion causes the measured penetration curve in Figure 21 to shift
toward larger particle diameters because Figure 10 is on the dry
particle size basis.
     2. System leaks - The actual scrubber penetrations could
also be different from those based on the  impactor data because
of dilution of the outlet by air leaks  into the system, which
probably did occur.  Dilution  of the outlet stream would cause
the actual penetrations to be  greater than those measured by a
factor equal to the dilution factor.  The  outlet concentration
may have been diluted as much  as 251.  The dashed line on Figure
21 includes the effects of 25% dilution and 1.5 times particle
growth on measured data.
     3. Collection mechanisms  - The  model  for venturi performance
assumes particle collection by inertial impaction on drops in
the throat region of the venturi.  This model does not account

                                55

-------
for forces other than inertia which can effect submicron par-
ticle collection.  Thus, in the region below about 0.5 ymA the
predicted penetration and the experimental penetration are expec
ted to differ.
     Another important phenomenon is the collection of particles
of all sizes due to solution-induced condensation.  Borax (Na2lU
is highly soluble in water.  If the scrubber liquor became more
concentrated than would be in equilibrium with the vapor, it is
conceivable that particles could be swept toward the liquid as
water vapor moves to condense on it.  The liquid drops which
contain dissolved borax have a reduced vapor pressure which in-
duces the condensation which causes the growth in particle size
manifested in the differences between wet and dry particle sizes.
     Some of this dissolution and condensation will have occurred
in the quench section upstream of the venturi.  However, because
of the large amount of  liquid injected at the venturi these
mechanisms may still be causing particle collection and growth
downstream of the venturi.  Growth or collection of the submicron
particles after  the inlet sampling point, either before or after
the venturi would explain some of the differences between the
predicted and the experimental penetrations.
     Condensation can also occur when the vapor pressure of the
water is reduced because it is at a lower temperature than the
adiabatic  saturation temperature of the gas.  This cause of con-
densation was not significant here because there was very little
cooling of the scrubber liquor.
     4. Sensitivity of prediction to L/G ratio - The predicted
curve is very sensitive to the liquid to gas volume flow rate
ratio (L/G) for particle diameters above 1.0 ymA.  A 25% decrease
in the L/G ratio would cause the prediction to agree closely with
the data.  Since neither the inlet gas flow rate nor the liquid
flow rate are known precisely, it is possible that the actual
L/G ratio was lower than the value used in the prediction.
     5. Entrainment - The carry-over from the cyclone was so heavy
that precutters had to be used on the outlet sampling probe.   It
                                 56

-------
was possible that some  smaller  entrapment  drops,  containing  dis-
solved borax, penetrated  the  precutters  or  were  shattered  in  the
precutters and were collected in the  heated impactors.   Since
the venturi model assumed that  no entrainment  carry-over occurred,
the actual penetrations would be greater than  the  model predictions
as seen in Figure 21.
     These five factors help  explain  the differences between  the
model prediction and  the  data.   The many factors and uncertainties
involved are enough to  preclude any judgment of  the accuracy  of
the model.

GAS-ATOMIZED SPRAY SCRUBBER COMPARISONS
     The gas-atomized spray scrubber  had a  throat  section  that
was a rectangular orifice, whose width could be varied to  adjust
 the pressure drop.
 Pressure Drop
     The geometry of  this scrubber was well known  except for  the
 "throat length."  Since the  throat was an orifice  the physical
 length was only a fraction of a centimeter. However, a free
 jet will not expand  immediately but will continue  straight for a
 few centimeters.  A  straight  section  length of 10  cm was chosen
 and as seen from Table  21 a  reasonable pressure  drop was predicted
 for some of the run  sets.  Predicted  pressure  drop for two of
 the run sets was very poor indicating either an  error in measure-
 ment or the assumed  straight  section  throat length was too large.
 Particle  Collection
      The  scrubber parameters are known or estimated  from  the
 average of the conditions for the sets of runs and shown  in
 Tables 21 and  22.   The predicted particle penetrations  are  shown
 compared  to  the  average of the run sets  in Figures 22-26.  The
 solid  (measurement)  curves represent the measured penetration with-
 out accounting for  leakage.   The dashed curves assume  that the
 outlet was diluted  by the factor listed in Table 22  which is
 based on  the  data  in Table 14.
      The  overall penetration predictions for the average  log-nor-
 mal inlet  dust distribution based on the cascade impactor data
 are shown  in  Table  23.
                                  57

-------
          TABLE 21.   PRESSURE DROP COMPARISON FOR THE
                     GAS ATOMIZED SPRAY SCRUBBER
Run Set uct
cm/s
5,6
8
9,
11,1
14,
Note:
,7 10,
23,
10 24,
2,13 10,
15 10,
Assumed
600
200
400
400
400
tt
QL
0.
0.
0.
0.
0.
= 10
/QG
0011
0009
0008
0031
0014
cm
ul Measured AP
de cm W.C.
0.
0.
0.
0.
0.

61
70
70
43
58

79
83
82
103
88

Predicted AP
cm W.C.
78
349
342
148
91

         TABLE 22.  CONDITIONS FOR GAS-ATOMIZED SPRAY
                    SCRUBBER PERFORMANCE PREDICTION
Run Set
5,6,7
8
9,10
11,12,13
14,15
PL
g/cm3
1
1
1
1
1
.0
.0
.0
.0
.0
PG
kg/m3
0
0
0
0
0
.72
.70
.36
.93
.80
2
2
4
1
2
yG
g/cm-s
.1x10"'
.2x10'"
.1x10'"
.7x10"*
.0x10"*
dd
ym
79
45
40
200
94

0
0
1
0
0
CDo
.68
.64
.10
.46
.61
B
2.26
2.00
2.02
7.30
2.86
Oil1
1.23
1.27
1.88
1.79
1.22
!Dil  is the ratio of measured outlet to inlet volume flow rates at
 standard conditions.
                               58

-------
         1,0
Cn
                                       WITH DILUTION FACTORH
                                                                              1.
        0.01
            0.1      0.2       0.5      1      2
                 PARTICLE AERODYNAMIC DIAMETER, d
                                                 pa'
 5
umA
10
       Figure  22.  Comparison of predicted with measured penetration
                   for average of runs 5, 6 and 7, for the gas-
                   atomized scrubber.
                                                                                                       WITH DILUTION FACTOR
                                                                             0.01
                  0.1    0.2       0.5      1      2         5      10
                           PARTICLE AERODYNAMIC DIAMETER, d  . pmA
                                                           pa
              Figure 23.  Comparison of predicted with measured penetra-
                          tion for run 8 for the gas-atomized scrubber.

-------
                                                                                1.0
g
        1.0
                                 WITH DILUTION  FACTOR
       0.01
0.1
0.2
                            0.5
                 PARTICLE AERODYNAMIC DIAMETER, d  ,  ymA
                                                 pa
       Figure 24.  Comparison of predicted with measured penetra-
                   tion for average of runs 9 and 10 for the gas-
                   atomized scrubber.
                                                                                             WITH DILUTION FACTOR
                                                                               .001
                                                                                        PARTICLE AERODYNAMIC DIAMETER,  d ft,  umA

                                                                              Figure 25.  Comparison of predicted with  measured penetra-
                                                                                          tion for average of runs 11,  12 and 13 for the
                                                                                          gas-atomized scrubber.

-------
    1.0
    0.5
H
U
o
I—I
H
 0.2
    0.1  :
0.05
        =; WITH DILUTION FACTOR
   0.02
   0.01

      0.1     0.2       0.5     1       2          5

             PARTICLE AERODYNAMIC  DIAMETER,  d   ,  ymA


   Figure  26.   Comparison of predicted with  measured penetra-

               tion for average of runs 14 and  15 for the gas

               atomized scrubber.
                          61

-------
          TABLE  23.  OVERALL PENETRATION COMPARISON FOR
                     THE GAS-ATOMIZED SPRAY SCRUBBER
Run Set
5,6,7
8
9,10
11,12,13
14,15
Measured *
Ft, %
13
5
5
7
15
Predicted
Ft, %
47
26
33
44
45
Note:  Assumed d   =0.4 ymA, a  =2.5
                ir o             o
* Neglecting outlet precutter dry weight gain.
Discussion
     The comparison of predicted and measured penetration leads
to the following observations:
     1. The predictions are generally close to the experimental
results for particles around 1  umA diameter.  Runs 5, 6, and 7
have the poorest fit.  Although the operating conditions for
these runs were similar to those of runs 14 and 15, the exper-
imental penetration was greater.  This may be due to the fact
that the gas flow rate was higher in runs 5,- 6, and 7 which could
mean more leakage and could cause more entrainment carryover at
the outlet.  The effect of L/G ratio on the model predictions
is shown by comparing the predicted penetrations of these two
run sets.  The main difference in the operating conditions was
that runs 14 and 15 had a 45% higher L/G ratio than runs 5, 6
and 7.
     2. Predicted penetration for sub-micron particles is gener-
ally higher than measured.  This may be due to condensation caused
by the cool duct walls and the cool scrubber liquid which had low
temperatures due to the cold ambient air (0°C).  Leaks of the
cold ambient air into the system would also have caused conden-
sation.
                                 62

-------
     3. Predicted penetration  is  generally  lower than measured
for particles larger than  about 1  ymA  diameter.  There are several
plausible causes of this disparity:
        a. Entrainment not collected in the entrainment separator
would yield particles in the larger size range.  The cyclone
separator was known to be  inefficient  although its performance
was improved by internal modifications and  leak patching.  This
inefficiency was clearly demonstrated  by noticeable liquid drops
at the outlet stack where  the  sampling crew was working, and by the
precutter mass collections.
        b. Imprecise knowledge of  the  liquid flow rate can cause
the prediction to be substantially incorrect in the larger par-
ticle size range.  This effect becomes more serious as the liquid/
gas ratio decreases, as seen by comparing the prediction of runs
4, 5, and 7 with that of runs  14  and 15.
     4. Predicted overall  penetration  was generally higher than
measured.  This difference is  probably due primarily to the
assumption that the size distribution  is log-normal with the same
standard deviation for all particles less than 0.4 ymA (dp ) .
Condensation probably affected the log-normality of the small
size distribution.  Also,  the  penetration model over-predicted
the penetration of the smaller particles which would definitely
cause the overall penetration  prediction to be too high.
     In conclusion, with all the  real  effects considered, the
model provides an adequate prediction.  Predictions for several
of the runs, notably 8, 11, 12, 13, 14, and 15, were quite good.
                                63

-------
                           SECTION 7

                     PERFORMANCE TEST METHOD

     The approach used in these scrubber evaluations involved
obtaining experimental collection efficiency data, acquiring
information on system characteristics and behavior, and per-
forming computations which made use of the performance data
and mathematical models.   Over the course of the program the
methods and apparatus used generally were improved and modified
to suit each specific test situation; however the main features
remained similar and are described here.
     The most important experimental measurements were those
regarding particle size and concentration.  Cascade impactors
were used to determine particle diameters larger than 0.3 ymA.
The Air Pollution Technology, Inc. portable screen diffusion
battery (A.P.T. - S.D.B.) was used for particle size determin-
ation in the diameter range from 0.01 ym to 0.1 ym (actual).
The apparatus and methods used are outlined below:
     1.  Gas velocity distribution and parameters were measured
at the inlet and outlet of the scrubber in order to define the
following:
         a. Conditio-ns required for isokinetic sampling.
         b. Particle mass concentration per unit volume of
dry gas, which is needed to provide a consistent basis for com-
paring inlet with outlet size distributions in the computation
of efficiency.
         c. Gas flow rate.
         d. Amount of liquid entrainment in the scrubber outlet
     Methods to measure these parameters are tabulated in the
table on the following page.
                                64

-------
       TABLE 24.  MEASURING  EQUIPMENT AND METHODS
Parameter                  Equipment
Gas velocity
and flow rate
Gas temperature
Humidity
                •••- -—•—•—- - . -   — _
Standard pitot tube or cali-
brated type "S" pitot tube;
differential pressure gauge.
Calibrated thermocouple or
mercury filled glass-bulb
thermometer.
Wet bulb - dry bulb
thermometers.
Method
 - • ' »-»^»^^«^»
EPA Method 1;
EPA Method 2.
Pressure
Inclined water manometer
or a pressure gauge.
Wet and dry bulb
temperature mea-
surement on a
flowing sample
withdrawn from
the duct.
Measured by means
of a static pres-
sure tube inser-
ted in the duct.
     2. The most essential part  of  a  scrubber performance
test is the determination of particle size distribution and
concentration  (loading)  in the inlet  and  outlet of the scrubber.
For accurate determination of particle size distribution, a
collection mechanism  that collects  particles and causes neither
formation nor breakup  of aggregates is necessary.  Cascade
impactors come close  to  meeting  these requirements.
     In a cascade  impactor particles  are  classified by inertial
impaction according to their mass.  The larger ones are collected
on the plate opposite  the first  stage and the smallest on the
plate opposite the last  stage.   A.P.T.  used the University of
Washington Mark III source test  cascade impactor for particle
size fractionation.   This impactor  classifies particles into
seven size ranges  and  is capable of sizing particles down to
about 0.1 ym diameter  (actual).   All  impactors were calibrated
in the laboratory  according to EPA  guidelines, see Calvert, et
al. (1976) and Harris  (1977) on  calibration method.
     In order to minimize probe  losses, tests were made with
the impactors in the  duct and with  the inlet nozzles appro-
                                65

-------
priately sized to give isokinetic sampling.  A modified EPA Method
5 sampling train was used to monitor the sample gas flow rate.
Figure 27 shows the sampling train arrangement.
     In some tests, a precutter was used to remove either the
large particles from inlet samples or the entrained liquid from
outlet samples.  A round jet impactor was found to work satis-
factorily as the precutter and was adopted for use for both inlet
and outlet sampling.  The impactors were either given time to
reach the duct gas temperature or heated to prevent condensation.
     To increase the weighing accuracy, lightweight substrates
were used on the collection plates. Generally, either greased
aluminum foil or a glass fiber filter paper substrate was used.
Impactor substrates and backup filters were weighed to the nearest
tenth milligram (10~"g) by using an analytical balance.
     Particle size distribution and loading measurements were con-
ducted simultaneously at the scrubber inlet and outlet.  This
minimizes the effects of particle size distribution changes which
may result from fluctuations in the operation parameters.  The
sampler was held at one location in the duct for the duration of
each sampling run.  This is an adequate technique for obtaining
good samples of particles smaller than a few microns in diameter
because they are generally well distributed across the duct.
     Blank impactor runs were performed periodically to assure
that the greased aluminum substrates did not react with the stack
gases or lose weight.  A blank impactor run consists of an im-
pactor preceded by two glass fiber filters and run at identical
sample conditions as the actual smapling runs.
     3.  In-stack filter samples were taken to obtain total par-
ticulate loadings and overall collection efficiencies of the
system.  The sampling train arrangement was the same as the cas-
cade impactor train (Figure 27) with the cascade impactor replaced
by a filter.
     4.  Inertial impaction devices such as cascade impactors are
normally insufficient for sizing particles smaller than 0.1 to
                                66

-------
PRECUTTER
   AND
 NOZZLE
          u
                                        THERMOMETER   IMpINGER TRAIN
                       HEATED
                       CASCADE
                      IMPACTOR
                    ORIFICE
                     METER
                                                                            MANOMETER
DRY GAS
 METER
VACUUM
 PUMP
                                      SILICA
                                      GEL
                                      DRYER
          Figure 27- Modified EPA sampling train with in-stack  cascade  impactor

-------
0.3 vim (actual diameter).  Size fractionation of these particles
may be accomplished by diffusional collection devices  (diffusion
batteries) which may consist of closely spaced parallel plates,
long, thin tubes, or an array of wire screens.  Parallel plate
and tube diffusion batteries require many pumps, dilution appara-
tus, and other bulky equipment which prove to be cumbersome in
field use.  For this and other reasons, Air Pollution Technology,
Inc. developed and uses a portable screen diffusion battery which
is lighter and more portable than previous devices.
     The screen diffusion battery utilizes a series of layered
screens intermittently separated for sampling purposes (Figure
28).  Size fractionation by the diffusion battery is detected by
measuring the overall particle number concentrations of the gas
stream entering and leaving a known number of screens.  A con-
densation nuclei counter (CNC) is used to determine number con-
centrations.  Concentrated aerosol samples must be diluted until
compatible with the CNC (~106 particles/cm3).  Screen penetration
data are then analyzed to determine size distribution and cumula-
tive mass loading of the particulates in the stream.  Cascade im-
pactor and diffusion battery analyses often can be combined to
obtain an overall characterization of the particulate size dis-
tribution (and scrubber penetration).
     Fine particle size measurements with the diffusion battery
were not taken simultaneously at the inlet and outlet of the
scrubber system.  During an impactor run, several inlet and out-
let measurements were taken alternately with the S.D.B.  Since
the system remained fairly constant during each run, alternate
inlet and outlet S.D.B. measurements were considered to approxi-
mate simultaneous sampling.
     Each S.D.B. run consisted of a continuous series of CNC
readings.  Normally, CNC counts were taken at each diffusion
battery stage in order of increasing number of screens and then
the process was repeated until three to four sets of readings
                                68

-------
  r
STACK
 GAS
                             1st DILUTION
                           FLASK WITH CHARGE
                             NEUTRALIZERS
             CASCADE
             IMPACTQR
                      ROTAMETERS
                         FILTER
                9
                                                               ROTAMETER
                                                                           VACUUM
                                                                            PUMP
          FILTER
               FLOW
               METER
                                                               1
                                        DIFFUSION
                                         BATTERY
                         2nd
                      DILUTION
                        FLASK
                                                                       TO
                                                                       CNC
                                                                 PUMP
DESSICANT
                                                                                        VACUUM
                                                                                  »« >  PUMP
               Figure  28,  Schematic diagram of diffusion battery system,

-------
were obtained.  Continuous monitoring of  flow,  temperatures,  and
pressures enabled steady operation of the diffusion battery.   Con-
ditions in the duct  (pressure, gas velocity, temperature,  and
water vapor content) were obtained during the impactor  tests.
     The size of particles entering the diffusion battery  was
limited by using a cascade impactor precutter on the  instack  end
of the probe.  loskinetic sampling was not maintained because
the particles to be measured were too small to  be segregated by
inertial effects from bends in the gas stream.  The sample stream
entering the diffusion battery was immediately  diluted with heated,
dried, filtered air to control condensation.  Two Polonium 210
charge neutralizers were inserted into the flask to eliminate
electrostatic effects.
     A portion of the resultant aerosol was directed through the
diffusion battery and the outlet was diluted to a concentration mea-
surable by the Gardner CNC.  The aerosol from the second dilution
flask was sampled with the Gardner CNC.  The excess aerosol was
exhausted through the vacuum pump.  The excess  aeros.ol from the
first dilution was passed through an absolute filter and pumped
to the atmosphere.  The Gardner CNC was calibrated daily against
a standard B.G.I. Pollack, Model P, CNC and found to read con-
sistently 33% lower than the Pollack CNC for the concentration
range used in these tests.

ERROR ANALYSIS
Sample Bias
     It is important to note that the program objective is to
investigate scrubber performance on fine particles and, conse-
quently, it is not essential that the methods used be accurate
for large particles.  This makes the sampling simpler in the
following ways:
     1.  Isokinetic conditions are not important for fine particles.
For example,  the error caused by sampling 4 ymA particles at a
velocity 50%  higher or lower than the gas stream velocity would
only be  about 2 or 3% of the concentration.

                               70

-------
     2. The fine particles  will be well distributed  in  the  gas
stream, except  in  cases  where  streams  with different particle
concentrations  have  not  had time to mix,  so that generally single
point sampling is sufficient.  To illustrate, we may  note  that the
Stokes stopping distance of a  3 ymA particle with  an initial
velocity of 15  m/s (50 ft/s) is about  0.04 cm (0.016")  and  for
a 1 ymA diameter particle it is one ninth of that.   Since the
stopping distance  is the maximum distance a particle can be dis-
placed from  a gas  stream line  by going around a right angle turn,
it  is  obvious that fine  particle distribution in the gas stream
will be negligibly affected by flow direction changes.
     3. The effect of a  precutter on the  size resolution of a
cascade impactor is  not  significant in the size  range of interest,
so  long as the  precutter has a cut diameter larger than several
microns.
Diffusion Battery
     The screen diffusion battery was  calibrated in  the A.P.T.
small particle  laboratory.   An aerosol of known  size distribution
was generated and  passed through the diffusion battery.  The
total number  concentration was measured with a condensation
nuclei counter at  the inlet and outlet of each S.D.B. stage.
The penetration (percent) of particles was then  calculated and
plotted against solidity factors on semi-logarithmic paper.  The
experiment was  repeated  with the same  aerosol until  a smoothed
average curve relating number  penetration to solidity factor
was obtained.   From  the  smoothed curve, a correlation factor was
found for computing  the  theoretical diffusion battery performance.
     The scatter of  data points about  the smoothed (fitted)
calibration curve  represents the experimental error  in  the  pene-
tration measurement.  This  measurement error included meter  reading
error and the accuracy of  the CNC.  The measurement  error was de-
fined in terms  of  relative  error, or the  deviation from the aver-
aged penetration value divided by the averaged value.
     This procedure  was  repeated on other aerosols of known
size distribution.  The  maximum relative  error was then determined
                                 71

-------
from these experiments for each solidity factor.  The maximum
relative error for the screen diffusion battery  determined by
this method  is 10.4% for  solidity  factors  of  13,  26,  and  40.
Cascade Impactors
     Cascade impactors were used as the principal means of  obtain-
ing information about the inlet-outlet size distributions.  It
was important to understand the sources of error and how  the
error can be minimized.
     The procedural errors include the accuracy of the weighing
of the deposits, and reading of the test data such as temperature,
gas volume,  time, and pressures.  The errors from the impactor
design and construction include wall losses, accuracy and pre-
cision in construction of critical components, and particle re-
entrainment  from the collection surface.
     Some of the design and construction limitations can be re-
duced by procedures such as recovering  the wall losses and by
sampling at  certain flow rates and times to reduce reentrainment
errors and by calibration of the impactor.   The experimental
data obtained with commercial impactors were reported by Smith,
et al. (1974).  Smith, et al. (1974) reported that all impactors
tested had appreciable wall losses for particle diameters above
10 microns.  This error can be reduced by brushing the material
from the wall onto appropriate collection disks.  The flow velo-
city through the impactor jets should not be above 65 m/s to be
absolutely certain of avoiding reentrainment of particles from
the collection substrates.  The extent of reentrainment will de-
pend on the properties of the material and the amount of deposit.
However,  Rao (1975) reported that collection efficiency increased
with increased particle load.  When the particle weight is over
10 mg, part of the deposit may break away from the surface and mi-
grate within the impactor.  The lightweight deposit places im-
portance on accurate weighing.  The analysis of impactor errors
was limited to the weighing error and in the calculation of
collection efficiency error.  The effects of weighing errors on
the results of impactor tests have been analyzed by Sparks  (1971),
                                72

-------
An analysis of the weighing  error using  three different estimations
was reported by Fegley,  et al.  (1975).   The results indicate that,
when the weight of sample per  stage  is less than 1 mg when weighed
with a balance with  a  precision of 0.05  mg, the error in the frac-
tional mass will be  greater  than 10%.
                                  73

-------
                            SECTION 8
              DATA REDUCTION AND COMPUTATION METHOD

CASCADE IMPACTOR DATA ANALYSIS
     In a cascade impactor particles are classified by inertial
impaction according to their mass.  The larger ones are collected
on the plate opposite the first stage and the smallest on the
plate opposite the last stage.
     Once the stage "catches" have been measured, usually by
weighing particle collection foils or papers, the data analysis
is relatively simple.  Generally the objective is to make a
plot of particle diameter versus mass percent oversize or under-
size and to represent the size distribution in terms of log-
normal distribution parameters if possible.  Thus, it is neces-
sary to do the following:
     1. Add all of the stage and filter collection weights to
get the total particle mass collected.
     2. Compute either:
        a. Cumulative percent collected as the gas flows through
           the succeeding stages.  This is "percent oversize."
        b. Cumulative percent penetrating as the gas flows
           through succeeding stages.   This is "percent under-
           size."
     3. Compute the cut diameters for the impactor stages,
taking into account gas viscosity (or temperature) and gas
sampling flow rate.  The equation is:
                           0,5
                                74

-------
where:   dpac = impactor  stage  cut  diameter,  ymA
           yG = gas viscosity,  poise
           d. = jet diameter, cm
           u. = jet velocity, cm/s

     The particle diameter  used is  called  "aerodynamic diameter"
and it has the unit of  "aerodynamic microns," ymA.  This is the
effective diameter for  particle separation by inertial impaction
and it takes into account the effects of particle density and
particle "slip" between gas molecules.  It  is related to the
actual physical size  of the particle by the following equation:
where:  d    =  aerodynamic particle  diameter, ymA
         pa
         d   =  actual particle  diameter, ym
         p   =  particle  density,  g/cm3
         C'  =  Cunningham slip  correction factor, dimensionless

     At room temperature for air the Cunningham slip correction
factor, C',  is given by:
                      C»  -  1  +  ----                    (9)
      If  the particle  distribution follows  the  log-normal law,
 a  straight line  will  result  on log-probability paper.  The 501
 value of particle  diameter  is  the geometric  mass mean diameter,
 d   ,  and the  geometric  standard deviation, a , is  given by:
 Pg                                          8
                       -  !i:**value of-^              (io)
                     g     50% value of d
                                        pa
                                 75

-------
Overall Particle Penetration
     Overall particle penetration is defined as:
                               y * dP                 cn>
            mass concentration out
            massconcentration in
where:     Pt = overall particle penetration, fraction
       Pt(d ) = penetration for particles with diameter
           "    d , fraction
        f(d ) = particle frequency distribution
         C  . = total inlet particle concentration, mg/DNm3

     The overall particle penetration can be computed using the
data from a simulatneous inlet and outlet cascade impactor or
filter run.
Particle Penetration as a Function of Particle Diameter
     The particle penetration for particle diameter, d ., or
grade penetration curve is given by:
                                    "dC
                          PtCdp)
                                      dC
                                            outlet
(12)
                                        P_
                                         \
                                            inlet
          dC
where:  —	*— = the slope of the cumulative mass concentration
        d(d )
           P     versus particle diameter curve at d ,
                 mg/DNm3- ymA

     Particle penetration as a function of particle size is com-
puted from inlet and outlet particle size distributions and
concentration data.  The major steps involved in the computation
are as follows:
                               76

-------
     1. Reduce cascade  impactor data to the form of cumulative
particle mass concentration  for each impactor cut diameter.
     2. Determine the slopes  of the cumulative mass distribution
curves at several values  of  particle diameter for both  the  inlet
and outlet and then  compute  penetration at each particle  diameter.
     There are several  techniques  to determine the slope  of the
cumulative mass  distribution curve.  Some  of the techniques are
discussed below.
     1. Graphical technique  - Cumulative mass concentration versus
aerodynamic particle diameter data may be  fitted with a curve
by eyeball method.   The slopes of  the curve at various  values of
particle diameter are then measured graphically.
     2. Curve fitting  - Curve fitting to the data points  and
the measurement  of curve  slopes by eye involves  subjective judg-
ment.  To eliminate  the judgment errors, it is possible to fit
the data with a  mathematical function and  then evaluate the slope
analytically.  We have  tried fitting the cumulative mass  curves
to log-normal distribution functions, and  the Weibull distribution.
Log-normal Distribution -
     If the  inlet and  outlet size  distributions  are nearly
log-normal,  then a purely mathematical particle penetration  is
used.  This  mathematical log-normal penetration is based  on the
following:
The cumulative distribution of particle sizes is:
                                                         (13)
 where   P(x)  is the cumulative mass fraction of sizes smaller
 than "d  ",  and
        pa '
                        In d   - In d
                    -\r "^  	  J»    | I,,   	  	 ~                  V.  s
                            lnag
                                 77

-------
The derivative of the distribution function is:
       d P  _ d P  d x   _     exp (-x2/2)              ,  ->
      * dpa " d x  d dpa "  C2iO* dpa in crg

thus,
      d C    C .  exp(-x2/2)
     =	E = -^	                            (16)
     d d     f 2ir)  d   In o
        pa          pa     g

Using equation (12) , the penetration  is then:

                                                 )
             :pti  Cln Vout
                              exp
Weibull Distribution -
     The Weibull distribution, Lipson and Sheth (1973), offers
two advantages over the log-probability distribution.  The first
is that it has three parameters rather than two.  The second is
that the cumulative distribution function (CDF) is explicit and
does not have to be approximated by multi-termed polynomials.

     Cumulative distribution function -
              CDF = 1 - exp
                                JE_lV\  |        -    (18)
                                6  " dpo
 where    d   =  particle diameter , cm
        d   =  minimum particle diameter, cm
          0  =  characteristic diameter ,  cm
          b  =  Weibull slope , dimensionless
                                78

-------
                  b  =  B
The CDF has  the  property that:





             CDF  (0 = d ) = 0.632                           (19)





The median particle diameter occurs when the CDF = 0.5,  so:






             V  • dpo + <9 - dpo) (in 2)1/b                (20)
    Linear transformation  - Transformation to a linear form:






             y = A + B x





 requires that:






            y = In  In  ( ^CDF)                              C22)




                            1                                (23)






 therefore,





            A - -b  In  (0 -dpQ)



                 B  = b
 and,
                                79

-------
     Least squares curve fit - The minimum particle diameter,
d  , is that which results in the highest linear correlation co-
efficient based on the above linear transformation, when a least
squares linear regression is performed on the size distribution
data.  Note that,

   0 < d   < smallest diameter found in the distribution
     -  po

     Density function - The Weibull density function is the de-
rivative of the CDF:
                    'd -d
                           ib-1
     f(dp) = ---i     exp
        p    e-d   \ e-d
                po \    poy
dP°
             (26)
     Penetration - The penetration is the ratio of the cumulative
mass loading distribution derivatives,

                 Pt (d )     to fout <
                           Cpti £in 
-------
"average"size of  the particles  in the distribution.   The median
particle diameter is directly related to "6" by equation (20).
     Weibull slope, physical  interpretation - "b" is analogous
to the geometric  standard deviation of the log-probability  dis-
tribution.  It  indicates  the  "spread" of the size distribution.
The larger the  Weibull  slope, bs  the more uniform (monodisperse)
the particle sizes.

SCREEN DIFFUSION  BATTERY  DATA ANALYSIS
     Screen diffusion battery data consist of particle number
concentrations  which are  obtained after  each stage in the diffu-
sion battery.   Particle penetrations are calculated  from the
ratio of the number concentration taken  at a given SDB stage to
the inlet number  concentration.   The particle size distribution
may then be determined  from the penetration data.
     Most particle formation  processes result in  a particle size
distribution which is log-normal.   Log-normal size distributions
are conveniently  represented  by two parameters:   the number geo-
metric mean diameter  (d  ) and the standard deviation (a ).
Typically, process created aerosols fail  log-normality only at
the extremes of large and small particles  which represent only
a small percentage of the total particulates.
     Calvert, et  al.  (1972) describe a method for converting log-
normal size distributions to  overall penetrations using the
relation between  particles of a discrete  diameter, and penetration
of those particles through the device.
     The equation is a  discrete form of  the following equation:

                     f
             Pt =  J  Pt f (dp) a~(ln dp)                 (28)
                   ~ OO
where, for a screen diffusion battery:
                                       B
                     Pt = exp (-A S dp )                 (29)
                                81

-------
and, Pt = penetration fraction of a particle of  a  given diameter
          through S
      S = solidity factor, a dimensionless parameter
     d  = particle diameter, ym
    A,B = constants established by theory and laboratory
          experimentation, dimensionless

and for a log-normal distribution:
    f(dp) -
             y/2/n In
                        exp
                     g
                                                2 -i
                                   dP - ln V
                                      In a
                                          g
(30)
     Calvert and Patterson (1977) have shown that for particles
     e size range of 10" um
through the SDB is given by:
in the size range of 10~3um < d  < 10"  um the penetration
      Pt = exp | -1 x 10"" S (us dr)"°"67d ~1>29
                              s  c-
                                                        (31)
where:  u  = superficial gas velocity, cm/s
        d  = wire diameter of screen, cm

Thus, the variables in equation (29) have the following values:

              A = 1 x 10"* (us dc)~0>67

              B = -1.29

     The particle diameter that penetrates a SDB stage with  501
efficiency is found by substituting 0.5 for Pt in equation  (29):
                                       1/B
                     pc
                           -In (0.5)
                              AS
                                                         (32)
                               82

-------
Each stage will have  a  certain "d  " for a given flow rate.
     In order to  determine  the number geometric  mean  diameter,
dpn, and standard deviation,  og,  of the particles entering the
SDB a curve matching  technique must be used.   The data consist
of overall penetration,  Pt  (ratio of outlet to inlet particle
concentration), and stage cut diameter, d  ,  for each  screen.
A plot of these data  points must  match a curve that is  a numeri-
cal solution  to equation (28) such  as one of  those presented
by Calvert, et al.  (1972).  The curve most closely matched will
determine the "d   " and "a  "  of the particles  entering  the SDB.
     Non-log  normal data must be  handled by a  graph stripping
technique outlined by Sinclair (1972) which entails tedious
graphical integration and mathematical conversions.  In our ex-
perience the  data have  fallen sufficiently close to log-normality
that the curve matching technique is acceptable.
     Conversion of number distribution to mass distribution is
necessary in  order to put the diffusion battery  and cascade
impactor data on  the  same basis.  The method used to make this
conversion  is a graphical integration of the  following  equation:

                N .     /-Npi dm     /N_ x 100
          c  . . _Bi    f   _P  3  [ _P	|         (33)
 /-
 f
J
            P1    100  J      dN     \    N
                     O         r      \

where:   N   = cumulative  number concentration  of particles
             smaller than "dp",  #/cm3
        N   = total number concentration of particles, #/cm3
         pt
         d   = particle diameter,  ym
        d P  = mass  of particles in the  infinitesimal  size range
         P1    (dp + 3dp),  g
        C .  = cumulative  mass concentration of particles
             smaller than "dpi", g/cm3
          83

-------
The quantity (dm /dN ) is simply the mass per particle of diameter
"d ".  The quantity (N  x 1QO/N  ) is the number percent of
particles smaller than ud ".  Thus, equation (33) can be evaluated
from a plot of mass per particle versus cumulative number percent
of particles, both quantities being evaluated at the same par-
ticle diameter to provide a point on the plot.   The total and
cumulative number concentration data are obtained as described
previously.
     For a log-normal distribution the mass and number distribu-
tion are related by:

               a  (number) = a  (mass)                  (34)
                o             o
            ln dPg = ln V + 3 ln
Finally, since impactor data are usually reported as mass aero-
dynamic impaction diameters,equation (8) must also be used to
complete the comparison of SDB and cascade impactor data.
                               84

-------
                          REFERENCES
Calvert, S. "Engineering Design of Fine Particle Scrubbers",
J. of APCA. 24: 929, 1974.

Calvert, S., C. Lake, and R. Parker "Cascade Impactor Calibra-
tion Guidelines" EPA-600/2-76-118, 1976a.

Calvert, S. , H.F. Barbarika, and C.F. Lake "National Dust
Collector Model 850 Variable Rod Module Venturi Scrubber
Evaluation"  EPA-600/2-76-282, 1976b.

Calvert, S., H.F. Barbarika, and G.M. Monahan  "Gas Atomized
Spray Scrubber Evaluation"  EPA-600/2-77-209a, 1977a.

Calvert, S., H.F. Barbarika, and G.M. Monahan  "American Air
Filter Venturi Evaluation"  EPA 600/2-77-209b, 1977b.


Calvert, S. and R.G. Patterson "Submicron Particle Size Measure-
ment with a Screen Diffusion Battery", Final Report, EPRI Con-
tract RP 723-1-760205, 1977.

Fegley, M.J., D.S. Ensor, and L.E. Sparks  "The Propagation of
Errors in Particle Size Distribution Measurements Performed
Using Cascade Impactors" Paper 75-32.5 presented at the 68th
Annual Meeting of APCA, Boston, MA, June 15-20, 1975.

Harris, D.B. "Procedures for Cascade Impactor Calibration and
Operation in Process Streams" EPA 600/2-77-004, January 1977.

Lipson, C.  and N.J. Sheth "Statistical Design and Analysis of
Engineering Experiments", McGraw-Hill, 1973.

Lundgren, D.A. "An Aerosol Sampler for Determination of Particle
Concentration as a Function of Size and Time", J. of APCA, 17:
225, 1967.

Rao, A.K. "Sampling and Analysis of Atmospheric Aerosols",
Particle Technology Laboratory, Mechanical Technology  Labor-
atory, University of Minnesota.  Publication  269, 1975.
                                85

-------
REFERENCES (continued)


Sinclair, D.   "A Portable Diffusion Battery" American Ind.
Hygiene Assoc. Journal 55: 729-755, 1972.

Smith, W.B.,  K.W. Gushing, and J.D. McCain "Particle Sizing
Techniques for Control Device Evaluations" EPA 650/2-74-102,
NTIS PB 240670/AS, October 1974.

Sparks, L.E.  Personal Communication, 1971.

Yung, S., S.  Calvert and H.F. Barbarika "Venturi Scrubber
Performance Model" EPA-600/2-77-172, 1977a.

Yung, S., H.F. Barbarika and S. Calvert "Pressure Loss in
Venturi Scrubbers" J. Air Poll. Control Assoc., 27: 548-551,
1977b.
                               86

-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
! REPORT NO. 2
EPA- 600/2 -78-032
4. TITLE AND SUBTITLE
Evaluation of Three Industrial Particulate Scrubbers
7. AUTHOR(S)
Seymour Calvert, Harry F. Barbarika, and
Gary M. Monahan
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Air Pollution Technology, Inc.
4901 Morena Boulevard, Suite 402
San Diego , California 92117
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
3. RECIPIENT'S ACCESSION- NO.
5. REPORT DATE
February 1978
6. PERFORMING ORGANIZATION CODE
8. PERFORMING ORGANIZATION REPORT NO.
	 1
10. PROGRAM ELEMENT NO.
1AB012; ROAP 21ADM-029
11. CONTRACT/GRANT NO.
68-02-1869
13. TYPE OF REPORT AJMD PERIOD COVERED 1
Final; 3/75-12/77
14. SPONSORING AGENCY CODE
EPA/600/13
is. SUPPLEMENTARY NOTESIERL_RTP project officer is Dale L. Harmon, Mail Drop 61, 919/ 1
514-2925. Earlier reports in this series are EPA- 600/2 -76 -2 82 and EPA-600/2-77-
209a and -209b. |
 16. ABSTRACT
 The report gives results of field measurements,  carried out on three full scale indus-
 trial scrubbers to determine scrubber performance characteristics, including par-
 ticle collection efficiency as  a function of particle diameter.  The three scrubbers
 were different gas-atomized  spray types with pressure drops ranging from 54 to 178
 cm W. C.  Their performance on major sources of fine particle emissions was com-
 pared to a mathematical performance model for venturi scrubbers.
17. KEY WORDS AND DOCUMENT ANALYSIS |
'• DESCRIPTORS
Pollution Atomizing
Dust Spraying
Scrubbers
industrial Processes
Measurement
Gases
18. DISTRIBUTION STATEMENT
Unlimited
b.lDENTIFIERS/OPEN ENDED TERMS
Pollution Control
Stationary Sources
Particulate
Collection Efficiency
19. SECURITY CLASS (This Report)
Unclassified
20. SECURITY CLASS (This page)
Unclassified
c. COSATI Field/Group I
13B
11G
07A
13H
14B
07D
21. NO. OF PAGES I
99
22. PRICE I
EPA Form 2220-1 (9-73)
                                        87

-------