EPA-650/4-74-026
June 1974
Environmental  Monitoring Series

         m
         :*£•

-------
                                  EPA-650/4-74-026
     COLLABORATIVE  STUDY
  OF METHOD  FOR  STACK  GAS
ANALYSIS AND  DETERMINATION
     OF MOISTURE  FRACTION
    WITH  USE  OF  METHOD  5
                   by

        Henry F. Hamil and Richard E. Thomas

           Southwest Research Institute
              8500 Culebra Road
            San Antonio, Texas  78284
             Contract No. 68-02-0626
               ROAP No. 26AAG
           Program Element No. 1HA327
        EPA Project Officer:  M. Rodney Midgett

  Quality Assurance and Environmental Monitoring Laboratory
        National Environmental Research Center
      Research Triangle Park, North Carolina 27711
                Prepared for

       OFFICE OF RESEARCH AND DEVELOPMENT
      U.S. ENVIRONMENTAL PROTECTION AGENCY
            WASHINGTON, D.C.  20460

                 June 1974

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

-------
                                 SUMMARY AND CONCLUSIONS
     This report presents the analyses of data which were obtained in the performance of EPA Method 3 (Gas Analysis
for Carbon Dioxide, Excess Air, and Dry Molecular Weight)  and in the determination of stack moisture content with
the use of EPA Method 5 for particulate matter determination. The data were obtained during collaborative testing of
Method 5.

     The collaborative tests were conducted at a Portland cement plant, a coal-fired power plant, and a municipal incin-
erator, using four sampling teams at each site. For this study, each sampling run at a test site is considered a repetition
at the same true level for both the stack gas analysis and the  moisture determination.  This assumption is made since
there were no independent methods for the determination of true values, and  since at a given site there were no signifi-
cant  changes in the determined values. At the cement plant, power plant, and  incinerator, there were 15, 16, and 12
runs, respectively. Not all collaborators completed  all runs, and thus there were missing values for the statistical analyses.
A total of 160 Method 3 determinations and 150 moisture determinations were submitted to statistical treatment.

     Precision estimates are obtained for the various parameters, with the exception of excess air, from an analysis of
variance based on a nested experimental design. These estimates are expressed in terms of within-laboratory, laboratory
bias, and between-laboratory components, and are presented below in terms of standard deviations.  Since the actual
gas composition undoubtedly varied slightly from run to run, with within-laboratory components contain source varia-
tions as well as sampling error and are probably larger than would be expected in the use of these analytical procedures
in the field.  The laboratory bias component is essentially free of this added variation due to the manner in which it is
calculated. The results obtained for each component are summarized below.

     Method 3- All collaborators used Orsat apparatus to perform their stack gas analyses. The average  of three con-
secutive analyses was used, but the requirement that they differ by no more than 0.2% by volume was not enforced.
There was no detectable CO at any of the test sites:

     (1)   CO-i . The within-laboratory standard deviation is estimated as 1 .44% C02 by volume, with 149 degrees of
           freedom. The laboratory bias standard  deviation is estimated as 1 .06% CO2 , with 9 degrees of freedom.  This
           gives a between-laboratory standard deviation of 1 .78% C02 . Particulate concentrations from compliance
           tests at municipal incinerators are corrected to 12% C02 . The demonstrated variation in a C02 determina-
           tion would cause the reported particulate concentrations of two laboratories who obtained the same
           uncorrected particulate concentrations to differ from each other by 36% at low C02 levels, and 16%
           at high CO2 levels.

     (2)   O2 . The within-laboratory standard deviation for O2 is estimated as 1.70% O2 by volume, with 149 degrees
           of freedom. The laboratory bias standard deviation is estimated as 1 .29% 02 , with 9 degrees of freedom.
           This results in an estimated between-laboratory standard deviation of 2.14% 02 .

     (3)   Dry Molecular Weight.  The within-laboratory standard deviation for dry molecular weight determination is
           0.20 Ib/lb-mole with 149 degrees of freedom. The estimated laboratory bias standard deviation is 0.14
           Ib/lb-mole with 9 degrees of freedom. From these, the between-laboratory standard deviation is estimated
           as 0.24 Ib/lb-mole. Thus, this determination is precise, even though there is considerable variation in the C02
           and O2 values used in the calculation.

     (4)   Excess Air. The excess air determination is shown to be a function of the O2  level, increasing exponentially
           as the % 02 increases.  The least squares estimation of the model is
           with a coefficient of determination, r2 , of 0.993. The equation was obtained using one quarter of the excess
           air values from the three sites.  From this model and the precision demonstrated for % 02 , normal deviation
           in % 02 determination can be expected to produce a variation of from 30% to 60% in the excess air value.
                                                     fli

-------
     Moisture Fraction—The within-laboratory standard deviation is estimated as 0.032 with 140 degrees of freedom.
The laboratory bias standard deviation estimate is 0.032 with 8 degrees of freedom. This gives a between-laboratory
standard deviation of 0.045.

     The following conclusions and recommendations are made based upon the results presented above:

     (1)   In a great deal of compliance testing, Method 3 is used only for the determination of the dry molecular
          weight of the stack gas, i.e., the CO2 and excess air values are not used in subsequent calculations. When
          this is the case, the requirement that 3 consecutive analyses differ by no more than 0.2% by volume may be
          relaxed. The precision shown for the dry molecular weight in this study without the restriction would be
          sufficient for field tests usage.

     (2)   When correction factors based upon the Orsat analysis are to be used, e.g. correction to 12% CO2 or correc-
          tion for excess air, it is imperative that the stack gas composition be  determined precisely.  Small variations
          in the C02 and 02 levels can produce relatively large variations in  these factors, and thus three consecutive
          analyses differing by no more than 0.2% by volume is a reasonable requirement.

     (3)   To allow more precise determination of stack gas composition, two relatively simple modifications of the
          standard Orsat gas analyzer could be made.

          (a)    The gas buret could be modified to allow direct reading to 0.1  mfi, with interpolation to the nearest
                0.05 mfi.

          (b)    A more accurate method of adjusting the pressure in the gas buret to atmospheric pressure could be
                installed.  The present hand-held leveling bulb could be replaced with a leveling bulb mounted in a
                screw-adjustable leveling clamp. Incorporation of a small sidearm manometer at the  top of the buret
                would allow precise adjustment of the pressure via the screw adjustment on the leveling clamp.  In-
                stallation of a stopcock in the sidearm manometer would be necessary to block off the manometer
                during filling of the gas buret and during transfer of the  gas to  and from the absorbing burets.
                                                    IV

-------
                                  TABLE OF CONTENTS

                                                                                        Page

LIST OF ILLUSTRATIONS	    w

LIST OF TABLES	    n

I.    INTRODUCTION	    1

II.   TEST DESCRIPTION	    2

     A.   Collaborative Test Sites	    2
     B.   Collaborators and Test Personnel	    2

III.   STATISTICAL DESIGN	    4

     A.   Terminology	    4
     B.   Experimental Design	    5
     C.   Data Handling and Analysis	    5

IV.   METHODS	    7

     A.   Carbon Dioxide	    7
     B.   Oxygen	    9
     C.   Dry Molecular Weight	   11
     D.   Excess Air	   12

V.   MOISTURE FRACTION	   15

APPENDIX A-Method 3. Gas Analysis for Carbon Dioxide, Excess Air, and Dry Molecular Weight   ...   17

APPENDIX B-Moisture Fraction Determination from Method 5	   21

APPENDIX C-Statistical Methods	   25

LIST OF REFERENCES	   37

-------
                                 LIST OF ILLUSTRATIONS



Figure                                                                                        Page




   1    Schematic Test Han, Method 3	    5




   2    Schematic Test Plan, Moisture Fraction	    5






                                      LIST OF TABLES




Table                                                                                         Page




   1    Carbon Dioxide Data	    8




   2    Oxygen Determination Data	    10




   3    Dry Molecular Weight Data	    11




   4    Percent Excess Air Data	    12




   5    Moisture Fraction Data	    16




 C.I    Analysis of Variance for % C02	    28




 C.2    Analysis of Variance for % 02	    30




 C.3    Analysis of Variance for Dry Molecular Weight	    31




 C.4    Analysis of Variance for Moisture Fraction	    34
                                                VI

-------
                                        I. INTRODUCTION
     This report describes the work performed on Contracts 68-02-0623 and 68-02-0626 and the results obtained on
Southwest Research Institute Project 01-3462-008, Contract 68-02-0626, which includes collaborative testing of the
method for stack gas analysis and the method for determination of stack gas moisture fraction with use of Method 5 for
particulate emissions as given in "Standards of Performance for New Stationary Sources"*1 \

     This report describes the statistical analysis of the data from collaborative tests conducted at a Portland cement
plant/2* a coal-fired power plant*3), and a municipal incinerator*4 ^

     The collaborative tests of the method for stack gas analysis and the method for determination of the stack gas
moisture fraction were not conducted as separate tests of Methods 3 and 4 O but as these methods are used in con-
junction with Method 5 for particulate emissions.

     The results of the data analyses  and the conclusions and recommendations based on these analyses are given in
this report.

-------
                                     II.  TEST DESCRIPTION
 A.   Collaborative Test Sites

      The site of the Portland cement plant test was the Lone Star Industries Portland Cement Plant in Houston, Texas.
 This plant utilizes the wet feed process and operates three kilns. The flue gas from each kiln passes through a separate
 electrostatic precipitator. The flue gases are then combined and feed into a 300-foot-high stack/2 *  Samples were taken
 at the sample ports located on the stack 150 feet above grade.

      Typical stack gas composition was about 7.5% C02 and 13.5% O2. No CO was detected. Moisture fraction (Bwo)
 was about 0.25.

      The site of the coal-fired power plant was the Allen King Power Plant, The Northern States Power Company, near
 St. Paul, Minnesota. The exhaust gas from the combustion chamber passes through the heat exchanger and splits into
 two identical streams upstream of twin electrostatic precipitators. The twin emission gas streams are fed into an 800-foot-
 high stack through two horizontal ducts/3)

      Samples were taken from sample ports located in the south horizontal duct upstream of the entrance to the stack
 flue. Typical stack gas composition was about 11.8% C02 and 6.4% O2. No CO was detected.  Moisture fraction (Bwo)
 was about 0.10.

      The site for the municipal incinerator test was the Holmes Road Incinerator, City of Houston, Houston, Texas.
 The facility consists of two independent parallel furnace trains.  Refuse feeds continuously onto traveling grate stokers in
 the furnaces. Gases leaving the furnaces are cooled in water spray chambers, and then enter the flue gas scrubbers to remove
 particulates. The gases are then drawn through induced draft fans and exhaust into the 148-foot-high stacks. Samples
 were taken from the sample ports located on the stacks 102 feet above grade. During the test, samples were taken from
 both units at the incinerator.  Typical stack gas composition was about 5.2% CO2  and 14.1% O2. No CO was detected.
Moisture  fraction (Bwo) was about 0.40.

      Stack gas samples were taken at all three sites during the performance of Method 5 determinations. Equal quan-
 tities of gas were taken at each traverse point to provide an integrated sample.  Stack gas was transferred from the stack
to a gas sample bag by means of a one-way squeeze bulb. Stack gas samples were analyzed by the Orsat procedure after
each day's runs.

      Moisture determination was made by the impinger method in conjunction with the Method 5 determinations.

 B.    Collaborators and Test Personnel

     The collaborators for the Lone Star Industries Portland Cement Plant test were Mr. Charles Rodriguez and
Mr. Nollie Swynnerton of Southwest Research Institute, San Antonio Laboratory, San Antonio, Texas; Mr. Mike Taylor
and Mr. Ron Hawkins of Southwest Research Institute, Houston Laboratory, Houston, Texas; Mr. Quirino  Wong,
Mr. Randy Creighton, and Mr. Vito Pacheco, Department of Public Health, City of Houston, Houston, Texas; and
Mr. Royce Alford, Mr. Ken Drummond, and Mr. Lynn Cochran of Southwestern Laboratories, Austin, Texas.

      The collaborators for the Allen King Power Plant test were Mr. Mike Taylor and Mr. Hubert Thompson of
 Southwest Research Institute, Houston Laboratory, Houston, Texas; Mr. Charles Rodriguez and Mr. Ron Hawkins of
 Southwest Research Institute, San Antonio Laboratory, San Antonio, Texas; Mr. Gilmore Sem, Mr. Vern Goetsch,
and Mr. Jerry Brazelli of Thermo-Systems, Inc, St. Paul, Minn.; and Mr. Roger Johnson and Mr. Harry Patel of Environ-
 mental Research Corporation, St. Paul, Minn.

     The collaborators for the Holmes Road Incinerator test were Mr. Mike Taylor and Mr. Rick Hohmann of Southwest
 Research  Institute, Houston Laboratory, Houston, Texas; Mr. Charles Rodriguez and Mr. Ron Hawkins of Southwest
 Research  Institute, San Antonio Laboratory, San Antonio, Texas; Mr. Quirino Wong, Mr. Randy Creighton, and

-------
Mr. Steve Byrd, City of Houston, Department of Public Health; Mr. John Key, Mr. James Draper, Mr. Tom McMicklc,
Mr. Tom Palmer, Mr. Michael Lee, and Mr. Charles Goerner, Air Pollution Control Services, Texas State Department of
Health.*

     The Portland cement plant test was conducted under the supervision of Dr. Henry Hamil, and the power plant
and municipal incinerator tests were conducted under the supervision of Mr. Nollie Swynnerton, both of Southwest
Research Institute.

     Collaborators for all three tests were selected by Dr. Hamil.
•"Throughout the remainder of this report, the collaborative laboratories are referred to by randomly assigned code numbers. For the
cement plant test, code numbers 101,102,103, and 104 are used. For the power plant test, code numbers 201, 202, 203, and 204 are
used. For the cement plant test, code numbers 301, 302, 303, and 304 are used. These numbers do not correspond to the above ordered
listing of laboratories. The ordering is the order that was used in the particulate collaborative studies. The first digit has been changed
to correspond to the site numbers used in this report.

-------
                                    III. STATISTICAL DESIGN
 A.   Terminology

      To facilitate the understanding of this report and the utilization of its findings, this section explains the statistical
 terms used in this report. The estimates of the pertinent values are developed in the subsequent sections.

      We say that an estimator, 0, is unbiased for a parameter 6 if the expected value of 0 is 6 , or in notational form ,
 E(d) = 6. From a population of method determinations made at the same true level, y., let xt , x2 , . . . ,xn be a sample
 of n replicates.  Then we define:

                  n
               1 ^-^
      (1)  x = — j ^ X( as the sample mean, an unbiased estimate of the true mean of the population of determina-
              n ~™
                i= i
          tions, 5. The sample mean gives an estimate of the center of the distribution of the determinations. If
          the method is accurate, 5 is equal to n, the true level.
     (2)  SS=     (Xi — x)2 as the sum of squares for the sample, which is used to estimate the dispersion of the
              i = 1
           population of determinations around 5.

     (3)   d/as the degrees of freedom, an indication of the amount of confidence in the estimate. A larger number
           of d/implies more confidence in the estimate.

     (4)   d2 = SS/df, as a variance estimate, or mean square, unbiased for a2 , the true variance of the determinations.
           The variance is a measure of the dispersion in the determinations around the true mean, 6 .
     (5)   a = vS2, as the estimated standard deviation of the determinations.  This term is a biased estimate of
           a = \Ar and is an alternative measure of dispersion.

     The variability in a method determination is expressed in terms of within-laboratory, laboratory bias and between-
laboratory components. The following definitions of these terms are given with respect to a true value, M.

     •     Within-laTjoratory—The within-laboratory component measures the dispersion in replicate single method
           determinations of the same true value, ju, made by one laboratory. The within-laboratory variance is
           estimated from the results of each laboratory at each test site and is denoted by a2 .

     •     Laboratory Was-The laboratory bias component measures the dispersion in determinations made of the
           same true value, p, due to use of the method by separate laboratories. These differences can be ascribed
           to such factors as different analysts and instrumentation, and the variance, a£, is estimated by comparing
           the results obtained by different laboratories at each test site.

     •     Between-laboratory— The between-laboratory component is estimated from the within-laboratory and
           laboratory bias terms.  The between-laboratory standard deviation is an estimate of the variation that can
           be expected between two single determinations made of the same true value, n, by two laboratories work-
           ing independently. The between-laboratory variance, a2, , is defined as

                                             a2=(i2+a2.

-------
B.    Experimental Design

      The data were collected from three separate tests of Method 5 at three sources covered by the new source per-
formance standards/1'  At each site, four collaborating laboratories were used, but the laboratories and collaborators
varied from one test site to another.  The number of sampling runs also varied from site to site.

     The model chosen is a nested or hierarchical moder ' with three factors: sites, labs-within-sites, and repetitions-
within-labs-within-sites or error. There were three sites; a cement plant, a power plant, and an incinerator.

      The determinations from each of the laboratories are considered only within the particular site where they were
made.  For each laboratory during each run of the test, determination of the stack gas composition by Method 3  and
moisture fraction determination by Method 5 were called for.

      For the analysis of the particulate matter determination at both the power plant and the cement plant, one labor-
atory's results were excluded. At the cement plant, Lab 102 deviated from Method 5 in the laboratory analysis of the
particulate matter. However, this had no bearing on either the Method 3 data or the moisture fraction data, and thus
Lab 102's results are included in this study. At the power plant, Lab 201 was eliminated due to the probable develop-
ment of leakage during  some  runs and filter contamination due to use of a low-melting ground joint lubricant.  Since
this would adversely affect the volumes of stack gas and liquid collected due to the introduction of ambient air into
the train, their moisture fractions are not usable. The Method 3 data from Lab 201 are unaffected and are included in
the analysis.

      The schematic of the design for the treatment of the various Method 3 results is shown in Figure 1.  The schematic
design for the analysis of the  moisture fraction data is shown in Figure 2.  The number of repetition, r/, varies slightly

                 SITE 1                           SITE 2                              SITE 3
       	I	        	I	         	I	
       I        I        I        I        I         I        I         I         I        I        T       I
   LAB 101 LAB 102  LAB 103  LAB 104 LAB 201  LAB 202  LAB 203  LAB 204 LAB 301 LAB 302  LAB 303  LAB 304
                    rl  rl4  r!   ri3 . -rl  rl6  r-i  ("is  /"i   r16  rj   r15  /^  r12  /"i  <"12

                               FIGURE 1. SCHEMATIC TEST PLAN, METHOD 3

                    SITE 1                        SITE 2                         SITE 3
          I                        I        I                 I         I                         I
      LAB 101 LAB 102 LAB 103 LAB 104  LAB 202 LAB 203  LAB 204  LAB 301  LAB 302 LAB 303  LAB 304
      r1  r14  r-i   /-15  r,  r14  r-t   r14  ^  r16  r,  r16  r,   r15   /^  r12  r,  r12  r,  r12  r,  /•-, ,

                          FIGURE 2. SCHEMATIC TEST PLAN, MOISTURE FRACTION

from lab to lab at each site due to a failure to complete the run, or in the case of Lab 303, failure to perform an
Orsat analysis.

C.    Data Handling and Analysis

      The raw data from the tests were used to obtain the determination values used in the analysis.  All Method 3
values shown were calculated using the three stack gas content determinations as a starting point, and the moisture
fractions were calculated from the dry gas volume and volume of liquid collected.

      Method 3 specifies that three consecutive analyses be made which differ by no more than 0.2 percent by volume
for each of C02 , 02 and CO. This requirement was not enforced in the collaborative tests due to time and  difficulty

-------
 factors.  It has been demonstrated* ^ that this requirement is stricter than can reasonably be expected by a qualified
 analyst using the specified equipment. Thus, the data as shown are the averages of three consecutive analyses on an
 integrated gas sample.

      The statistical model for this experiment is of the form
where

                                 Yjjic is the fcth repetition by lab; at site /.

                                 IJL is the overall mean.

                                 •Yt is the effect of the i*h site.

                                 tyl/ is the effect due to laboratory; at site i.

                                      is the random error associated with Yiik.
      The site factor is not of interest, since it merely reflects the differences in the levels of the parameters of
interest from site to site.  Its inclusion in the analysis serves as a restriction on the error term by removing
these effects.

      The lab factor, X/|/, provides an estimate of the laboratory bias variance by comparing the results from different
laboratories at the same site.  The error term, ejtl/lj, is the source of the within-laboratory variance, assumed constant,
and comes from comparison of results by the same laboratory at the same site.

      The sampling runs at each site are considered replicate determinations of the same true level for the various factors
studied. This is done since the true levels were unknown, no independent means were available to measure them, and
since inspection of the data does not indicate a great disparity in the level of any factor. As a result, the error term
reflects both normal  sampling error and fluctuation in the true level and is probably larger than the true within-labora-
tory variance. The laboratory bias term is essentially free from any error due to level fluctuation, since it is determined
by comparing the averages of all the runs.

-------
                                          IV.  METHODS
     Method 3 is for the determination of C02, dry molecular weight, and percentage of excess air. The method calls
for the use of an Orsat analyzer or equivalent to determine the CO2,O2 , and CO content of the stack gas. All the
analyses in this study were performed using Orsat equipment.  At all sites tested, there was no detectable percentage
of CO, and thus the resultant variables for study with regard to Method 3 were:

     (1)   %C02.

     (2)   %02.

     (3)   Dry Molecular Weight (Md.).

     (4)   Excess Air (%EA).

These variables were considered both with respect to the precision that can be expected in their determination and,
where  applicable, to the degree that their imprecision could affect the results of a performance test for compliance.
The results of the statistical treatment are presented in the following sections, while more detail of the analyses is
contained in the appropriate Appendix C section.

A.    Carbon Dioxide

      The C02 determinations made by the collaborators for the three test sites are shown in Table 1. These values
were used in an Analysis of Variance (AOV) on the nested design to give the following results in terms of the precision
associated with a single C02 determination by Method 3 using an Orsat analyzer. The precision estimates are
Obtained in Appendix C.2.

      The within-laboratory variance estimate for the % CO2 determinations is

                                               a2  = 2.06

This estimate has 149 degrees  of freedom associated with it. The estimated within-laboratory standard deviation is
given by
                                                 = V2^06

                                                 = 1.44%CO2

      The laboratory bias variance, a\, is estimated as



 with 9 degrees of freedom. The estimated laboratory bias standard deviation is
                                                   = 1.06%CO2.

-------
TABLE 1. CARBON DIOXIDE DATA
       %CO, by volume

Run

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

Run

1
2
3
4
' 5
6
7
8
9
10
11
12
13
14
15
16

Run

1
2
3
4
5
6
7
8
9
10
11
12
Sitel
Labs
101
9.0
7.4
6.0
7.0
7.0
10.0
4.6
4.6
6.5
5.8
6.8
7.0
6.0
_*
4.7
102
9.0
9.0
9.0
7.0
10.4
10.5
4.0
7.2
7.6
8.0
8.0
7.6
9.6
9.6
8.6
103
_*
9.0
9.2
11.4
9.1
9.1
5.4
5.8
7.1
7.6
7.2
7.2
6.6
8.5
8.3
104
2.9
7.2
5.1
8.3
4.0
10.0
4.6
_*
7.5
7.4
7.2
6.7
_*
8.1
6.4
Site 2
Labs
201
7.6
7.6
9.4
9.7
8.7
10.3
9.9
9.8
9.7
10.0
8.1
10.2
12.8
13.1
11.7
12.4
202
5.2
8.0
12.6
12.0
13.0
12.6
10.5
11.5
13.2
13.6
12.7
12.2
13.1
12.4
10.6
10.2
203
13.4
13.3
12.7
12.6
12.9
12.7
13.1
13.7
12.9
13.4
12.9
12.3
12.8
13.0
13.4
13.1
204
13.2
12.1
11.6
13.2
13.0
12.5
12.4
12.6
11.9
12.8
12.7
12.4
12.5
_*
13.0
12.6
Site 3
Labs
301
6.4
4.3
6.2
6.0
7.2
5.2
5.7
5.0
5.1
5.5
6.1
5.7
302
4.8
4.0
4.3
6.8
7.1
5.6
6.4
6.1
6.6
5.9
7.9
4.9
303
4.8
4.0
1.0
5.0
5.0
5.2
6.1
_*
-t
-t
-t
-t
304
_*
3.8
3.0
3.7
3.7
3.2
4.1
5.4
5.3
4.3
4.8
3.0
*Run aborted, no Great data taken.
fNo analyses made.
     Combining the previous estimates, the between-laboratory
variance, oj , is estimated as
                                                                 = 2.06 + 1.12

                                                                 = 3.18.

                                       This gives an estimated between-laboratory standard deviation of
                                                                   = 1.78%C02

                                             The percent C02 determination is used as a correction factor
                                        in the determination of particulate emissions from incinerators accord-
                                        ing to the formula
                                                                     12
                                                                   %C02
                                        The effect of a deviation from the actual %CO2 upon this correction
                                        factor and upon the particulate concentration determination can
                                        be demonstrated by considering the case where the determined
                                        value differs from the actual by one standard deviation. The between-
                                        laboratory standard deviation is used since this gives an indication
                                        of the comparative results of two independent testing facilities
                                        working at the same true particulate concentration and C02 level.
                                        Let,

                                                F,- be the corrected concentration value for lab i

                                                C be the uncorrected concentration
                                                 j be the correction factor determined by lab /.
                                        Then
                                             Suppose that one laboratory determines the correct CO2
                                        percent and the other differs by one between-laboratory standard
                                        deviation. Then the factors are
                                                                       12
                                                                     %CO2
                                                             k        12
                                                                   %CQ2±ob'

                                        and the concentrations are reported as

-------
                                              /     12
                                          2  ~ \%C02 ±

      The error that would be induced by a one standard deviation error can be shown by taking the ratio of the
two concentration values, V\ / F2.

                                                 12
                                               %CO2
                                                   12
                                               % C02  ± ofc

                                               12C     %CO2±ab
                                             ~%C02        12C

                                             _ % CO2 ± Qfr
                                                %CO2

      Assuming true values of 5, 8, and 1 1 percent for % CO2 , the resultant error can be demonstrated. For 5 per-
cent, one standard deviation high gives
                                           V2        5%

                                                6.78
                                                 5

                                              = 1.36.

One standard deviation low gives
                                                 5% -1.78%
                                                     5%

                                                3.22
                                              = 0.64.

Similarly, for true values of 8 and 11 percent the ratios are 1.22 and 0.78, and 1.16 and 0.84, respectively. Thus,
the data from these tests indicate that variation of one standard deviation in the CO2 level would cause the reported
particulate concentrations of two laboratories to differ by 36 percent at low C02 levels, and 16 percent at high CO2
levels, when corrected to 12% CO2.

B.   Oxygen

     The percentage of oxygen in the stack gas is also measured by the Orsat analyzer. There is no direct application
of this to a standard of performance, but it is used in the computation of both the dry molecular weight of the gas
stream and the percentage of excess air. The 02  determinations from the three sites are presented in Table 2.

-------
TABLE 2. OXYGEN DETERMINATION DATA
           % O, by volume

Run

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

Run

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

Run

1
2
3
4
5
6
7
8
9
10
11
12
Sitel
Labs
101
12.0
13.6
14.2
13.7
13.7
10.7
16.2
16.2
13.2
14.1
13.9
13.7
14.2
_*
14.7
102
L2-2
12.2
12.8
14.8
11.4
11.5
17.0
14.4
14.0
14.0
13.2
14.0
12.2
12.6
13.0
103
_*
15.4
11.5
10.6
11-7
11.4
16.0
14.2
13.0
13.6
13.4
13.8
14.2
12.3
12.3
104
16.8
6.1
16.5
13.4
17.8
11.2
16.2
_*
13.8
12.7
13.9
14.5
_*
13.0
14.8
Site 2
Labs
201
8.1
8.1
8.2
8.2
9.3
7.3
8.0
8.1
8.0
7.8
9.0
7.0
4.3
4.8
4.7
4.2
202
14.3
11.4
5.3
6.7
5.2
6.2
7.0
6.1
5.1
4.9
5.8
6.3
6.3
6.5
7.3
8.2
203
4.6
5.0
5.5
5.9
5.3
5.4
5.0
4.6
5.5
5.4
5.0
6.0
6.0
5.0
5.0
5.0
204
4,8
6.4
6.4
5.2
5.0
6.0
5.8
5.9
6.4
5.6
5.8
6.2
5.8
_*
5.5
5.8
Site3
Labs
301
14.0
15.3
13.7
13.7
13.5
14.7
14.1
14.5
14.6
14.3
13.8
14.2
302
14.2
14.0
13.4
10.8
10.3
11.0
10.3
10.0
10.3
10.2
8.6
10.2
303
15.0
11.5
20.0
15.0
15.0
14.9
14.1
-t
-t
-t
-t
-t
304
_*
15.7
16.7
20.0
17.0
16.8
16.7
14.8
14.1
15.3
16.6
17.5
*Run not made.
t Orsat data not taken.
     In Appendix C.3, the AOV table is shown for these deter-
minations and the appropriate variance components estimated.
Under the assumption that the 02 level remained essentially con-
stant over the testing period, we can estimate a within-laboratory
variance, a2 , of

                         a2 = 2.90

with 149 degrees of freedom. This gives an estimated within-
laboratory standard deviation of
                                                                     = \/2~90

                                                                     = 1 .70% 02 by volume.

                                                From the laboratory factor of the analysis, the laboratory
                                          bias variance, o\ , may be estimated as

                                                                   &L = 1.66

                                          with 9 degrees of freedom. Then the estimated laboratory bias
                                          standard deviation is
                                                                    = 1.29% O2 by volume.

                                                Combining the above components, we can estimate a between-
                                          laboratory variance, a|, by the formula in section IIIA. The estimated
                                          value is
                                                                dl = (1.66) + (2.90)

                                                                   = 4.56

                                          Then the between-laboratory standard deviation is estimated by
                                                                     = 2.14%02 by volume.

                                                Thus, the between-laboratory standard deviations for both
                                          C02 and 02 are in fairly close agreement. This is consistent with
                                          comments made by users of the Orsat method that there is a trade-
                                          off between COZ and 02 . That is, a loss of CO2 results in an
                                          equivalent gain in 02 .
                                                10

-------
TABLE 3. DRY MOLECULAR WEIGHT DATA
               Ib/lb-mole

Run

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

Run

1
2
3
4
"5
6
7
8
9
10
11
12
13
14
15
16

Run

1
2
3
4
5
6
7
8
9
10
11
12
Site 1
Labs
101
29.92
30.01
29.53
29.67
20.67
30.03
29.38
29.38
29.57
29.49
29.64
29.67
29.53
_
29.34
102
29.93
29.93
29.95
29.71
30.12
30.14
29.32
29.73
29.78
29.84
29.81
29.78
30.02
30.15
29.90
103
_*
30.06
29.93
30.53
29.92
29.91
29.50
29.22
29.38
29.76
29.69
29.70
29.62
29.85
29.82
104
29.14
29.40
29.48
29.86
29.35
30.05
29.38
_*
29.75
29.41
29.71
29.65
_*
29.82
29.62
Site 2
Labs
201
29.54
29.54
29.83
29.96
29.76
29.94
29.90
29.89
29.87
29.91
29.66
29.91
30.22
30.29
30.06
30.15
202
29.40
29.74
30.23
30.19
30.29
30.26
29.96
30.08
30.32
30.37
30.26
30.20
30.35
30.24
29.99
29.96
203
30.33
30.33
30.25
30.25
30.28
30.25
30.30
30.38
30.28
30.36
30.26
30.21
30.29
30.28
30.34
30.30
204
30.30
30.19
30.11
30.04
30.28
30.24
30.22
30.25
30.16
30.27
30.26
30.23
30.23
—
30.30
30.25
Site3
Labs
301
29.58
29.30
29.54
29.51
29.69
29.48
29.48
29.38
29.40
29.45
29.53
29.48
302
29.34
29.20
29.22
29.52
29.27
29.34
29.44
29.38
29.47
29.35
29.61
29.19
303
29.65
29.10
28.96
29.40
29.40
29.43
29.54
-t
-t
-t
-t
-t
J04
_
29
29.24
29.15
29.27
29.18
29.32
29.46
29.41
29.30
29.43
29.18
*Run not made.
fNo Orsat data taken.
Note: EPA policy is to express all measurements in
Agency documents in metric units. When
implementing this practice will result in undue cost or
difficulty in clarity, NERC/RTP is providing
conversion factors for the particular non-metric units
used in the document. For this report, the factor is:
1 Ib/lb-mole o 1 gm/gm-mole.
C.    Dry Molecular Weight

      The dry molecular weight (Ma) of the stack gas is determined
from the stack gas analysis. The formula is

  Md = (0.44) % C02 + (0.32) % O2 + (0.28) (%N i + % CO).

The sites tested had no detectable CO, and since the percent N2
is determined by subtraction, the values for Md used in this report
depend solely on theC02 and 02 determinations. Thus, in this
section the precision of the Ma determination is given, along with
the relationship of that precision to the precision of the C02 and
02 determinations.

     The dry molecular weights used in the analysis are shown in
Table 3.  Submitting these  to an AOV according to the model
discussed gives the precision estimates desired. The values are
obtained in Appendix C.4.

     The within-laboratory variance, o2 , is estimated as

                       o2 = 0.04

with 149 degrees of freedom. From this, the estimated within-
laboratory standard deviation is
                                                                   = 0.2 Ib/lb-mole.

                                                    The estimated laboratory bias variance is

                                                                     al = 0.02

                                               with 9 degrees of freedom. Thus, the estimated laboratory bias
                                               standard deviation is

                                                                    OL = V0.02

                                                                       = 0.14 Ib/lb-mole.


                                                    Combining estimates, the between-laboratory variance, a2,,
                                               for the determination of dry molecular weight is estimated as
                                                                      = (0.02) + (0.04)

                                                                      = 0.06,

                                               and the estimated between-laboratory standard deviation is

                                                                    ab = -v/0.06

                                                                       = 0.24 Ib/lb-mole.
                                                    11

-------
TABLE 4. PERCENT EXCESS AIR DATA

Run

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

Run

1
2
3
4
5
6
7
8
9
10
11
12
13
14
IS
16

Run

1
2
3
4
5
6
7
8
9
10
11
12
Site 1
Labs
101
135.5
180.9
206.8
189.4
189.4
104.5
344.0
344.0
165.0
200.1
197.6
189.4
206.8
_*
223.5
102
141.8
141.8
163.2
253.2
123.3
126.5
440.9
228.7
209.0
212.4
173.6
209.0
144.5
156.6
168.9
103
_*
337.8
121.9
103.4
127.1
118.9
336.8
213.3
166.0
188.8
177.2
195.6
211.7
142.9
142.0
104
381.9
36.3
393.1
184.3
625.7
116.6
344.0
_»
197.8
156.2
200.6
230.0
_*
166.0
246.5
Site 2
Labs
201
57.2
57.2
60.5
60.9
75.3
50.5
58.5
59.7
58.3
56.1
69.8
47.1
24.5
28.4
27.1
23.6
202
205.7
115.4
32.4
45.4
31.7
40.7
47.4
39.0
31.0
29.5
36.9
41.4
42.1
43.6
50.8
61.5
203
27.0
30.2
34.2
37.8
32.5
33.3
30.1
27.1
34.3
33.7
30.0
38.5
38.9
30.0
30.0
30.1
204
28.5
42.3
42.0
32.3
30.0
38.7
36.7
37.8
42.2
35.1
36.9
40.6
36.8
_*
34.3
36.8
Site 3
Labs
301
199.6
258.2
184.0
182.7
181.6
226.2
199.4
214.8
221.2
208.1
187.8
204.4
302
197.7
183.1
160.9
98.6
91.6
99.8
88.1
82.3
88.5
85.4
64.0
83.5
303
233.0
106.4
2336.4
245.1
245.1
240.6
202.4
-t
-t
-t
-t
-t
304
	
282.8
371.2
13966.5
432.0
388.9
396.8
236.2
196.4
258.2
400.0
501.7
*Runs not made.
|No Orsat data taken.
     Thus, there is little variation in the computed dry molecular
weights, especially in light of the variation demonstrated previously
for the CO2 and O2 determinations. This has been noted in
other studies on the Orsat method/7 >  The trade-off between C02
and 02 mentioned earlier is responsible for this small deviation
in Md.  If we assume a loss of 2% CO2  by volume and a result-
ant gain of 2% 02 by volume, the net effect on Md would be:

   Ma = (0.44) (% C02 - 2%) + (0.32) (% O, + 2%)
        + (0.28)(%N2 +%CO)

      = (0.44) (% C02)- (0.44) (2%) + (0.32) ($O2) + (0.32)(2%)
        + (0.28)(%N2 +%CO)

      = (0.44) (%C02 ) + (0.32) (%02 ) + (0.28) (%N2 + % CO)
        + (032) (2%) - (0.44) (2%)

      = dry molecular weight + (0.32 - 0.44) 2%

      =  dry molecular weight - 0.24.

The result would be  an estimate that fell 0.24 below the actual
value. A comparison of the between-laboratory standard
deviations for CC^ ,02 and Md gives credence to this trade-off
explanation.

D.   Excess Air

      The excess air percentage is determined according to the
 formula
                                            %£A =
                                                            (%02)-0.5(%CO)
                                                   (0.264)(%N2 ) -
                                             X 100 percent
                                            Using the data from Tables 1 and 2, the excess air deter-
                                            minations were computed, and the values are shown in
                                            Table 4.
                                                 It is apparent from the data that there is a great deal of
                                            imprecision in this computation. Rather than apply an AOV
                                            approach to obtain precision estimates, the factor or factors
                                            upon which the %EA determination depends is investigated.
                                            In the absence of CO, the above formula can be expressed as
                                                   %EA =
                                                                   %02
                                                          (0.264)(%N2)-%0,
                                    X 100 percent
                                            and since %N 2 = 100 - % O2 - % CO2,
                                                %EA =
                                                                    %02
                                                       (0.264X100- %02 - %CO2)- %02

                                                       	%02	
                                                      " 26.4- (1.264)%02 - (0.264)% C02
                                               X 100 percent


                                              X 100 percent.
                                               12

-------
Thus, the excess air is the ratio of the oxygen content to a constant less a percentage of the 02 and C02 fractions.
But since only about one-quarter of the % CO2 is involved in the denominator, this amounts usually to a relatively
small contribution to the overall determination.  As a result, the factor making the significant contribution to the
estimate is the oxygen percentage.

      Based upon this and the nature of the data, an exponential model for the excess air was proposed as
where

                                               y is % EA

                                               x is % O2

                                   e is the base of the natural logarithms

and

                                           a0,a\ are constants.

      A least squares fit of a sample of the excess air determinations taken from all three sites was used to estimate
a0 and a\  and to determine the degree of fit. The steps used are shown in Appendix C.5. The model obtained was
The degree of fit, as measured by the coefficient of determination, r2 , is

                                               r2 = 0.993

indicating an extreme closeness of the determinations to the model .

      By using this model and the previously determined precision estimate for % 02 determination, it is possible to
demonstrate the imprecision of the excess air determination. Using the between-laboratory standard deviation of
2.14% 02 , we can estimate the effect on the excess air determination of an error of one standard deviation in the 02
determination.
      If we let x = % O2 be the actual percentage of O2 in the gas, and y be the actual percentage of excess air, then
a standard deviation of 2.14% O2 implies
where yc is the calculated excess air. Then

                                 yc = (10.47)e <°-

                                    = [(10.47)e (°-
                                                    13

-------
Thus a normal error in 02 determination would result in a 57% error in the excess air determination.  Similarly, a
one standard deviation error in the negative direction implies
                                                  = (0.64)y

      From the above, it can be seen that the excess air determination can only be as reliable as the O2 determination,
and due to the exponential relationship, small deviations in 02 result in large deviations in excess air.
                                                      14

-------
                                   V. MOISTURE FRACTION


     The moisture fractions for this study were obtained using the formula given in Method 5. The equation is
where

     Vw td-the volume of water vapor collected, corrected to standard conditions.

     Vm   .— the volume of gas collected, dry basis, corrected to standard conditions.

The water vapor collected is determined by adding the water volume increase in the impingers to the weight increase in
the silica gel tube in grams, and multiplying by a constant factor. Using the above formula, the moisture fractions used
in the analysis were obtained, and these are shown in Table 5.

     The precision estimates for the moisture fraction determination are obtained using an AOV in the design that was
shown in Figure 2. The AOV table and related data are shown in Appendix C.6.

     The within-laboratory variance, a2 , is estimated as

                                              &2 = 0.001

with 140 degrees of freedom. This gives an estimated within-laboratory standard deviation of
                                                = 0.032

     The laboratory bias variance estimate obtained was

                                              Of, = 0.001

with 8 degrees of freedom. The laboratory bias standard deviation, then, is given by
                                                 = 0.032

     Using the above values, the estimated between-laboratory variance is

                                        6« = &i + d3

                                           = (0.001) + (0.001)

                                           = 0.002.

and the between-laboratory standard deviation is estimated as
                                                = 0.045
                                                   15

-------
TABLE 5. MOISTURE FRACTION DATA

Run

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

Run

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

Run

1
2
3
4
5
6
7
8
9
10
11
12
Sitel
Labs
101
0.26
0.30
0.30
0.32
0.31
0.31
0.20
0.22
0.23
0.23
0.25
0.26
0.26
_*
0.26
102
0.33
0.31
0.33
0.30
0.38
0.32
0.21
0.23
0.26
0.26
0.28
0.28
0.28
0.30
0.28
103
_*
0.19
0.24
0.27
0.28
0.26
0.21
0.22
0.18
0.22
0.23
0.21
0.22
0.24
0.23
104
0.28
0.27
0.23
0.30
0.32
0.30
0.16
0.22
0.25
0.24
0.28
0.28
_*
0.29
0.28
Site 2
Labs
202
0.10
0.10
0.10
0.10
0.11
0.10
0.11
0.10
0.10
0.10
0.11
0.11
0.10
0.10
0.11
0.10
203
0.05
0.09
0.09
0.10
0.09
0.10
0.08
0.09
0.09
0.09
0.10
0.09
0.09
0.10
0.10
0.10
204
0.10
0.08
0.09
0.10
0.10
0.10
0.10
0.10
0.09
0.09
0.11
0.10
0.10
_*
0.10
0.10
SiteS
Labs
301
0.37
0.37
0.30
0.38
0.42
0.35
0.37
0.38
0.35
0.32
0.35
0.38
302
0.29
0.27
0.30
0.32
0.33
0.29
0.32
0.31
0.32
0.29
0.32
0.28
303
0.37
0.43
0.41
0.44
0.44
0.40
0.41
0.38
0.43
0.41
0.43
0.39
304
_*
0.40
0.40
0.23
0.32
0.37
0.40
0.42
0.43
0.40
0.42
0.36
*Runs not made.
                 16

-------
              APPENDIX A

METHOD 3. GAS ANALYSIS FOR CARBON DIOXIDE,
  EXCESS AIR, AND DRY MOLECULAR WEIGHT.
                    17

-------
24886
                                                  RULES  AND REGULATIONS
MKTHOD .1—GAS ANALYSIS FOR CAHBON DIOXIDK,
  KXCKSS AIR. AND  CUT MOLECOLAJt WEIGHT

  1  Principle and applicability.
  1.1  Principle. An Integrated  or  grab  gas
sample is extracted from a sampling point
and  analyzed for  Its components using an
Orsat analyzer.
  1.2  Applicability. This  method should be
applied only when specified by the test pro-
cedures for determining compliance with the
New Source Performance Standards. The test
procedure will Indicate whether  a grab sam-
ple or an integrated sample Is to be used.
  2.  Apparatus.
  2.1  Grab sample (Figure 3-1).
  2.1.1  Probe—Stainless  steel   or  Fyrex>
glass, equipped with a filter to remove partic-
ulate matter.
  2.1.2  Pump—One-way  squeeze bulb, or
equivalent,  to  transport gas  sample  to
analyzer.
  1 Trade name.
                                             2.2  Integrated sample (Figure 3-2).
                                             2.2.1  Probe—Stainless   steel  or  Pyrez1
                                           glass, equipped with a filter to remove par-
                                           ticulate matter.
                                             2.2.2  Air-cooled condenser or equivalent—•
                                           To remove any excess moisture.
                                             2.2.3  Needle valve—To  adjust flow rate.
                                             2.2.4  Pump—Leak-free,  diaphragm  type,
                                           or equivalent, to pull gas.
                                             2.2.5  Rate meter—To  measure  a now
                                           range from 0 to 0.035 cfm.
                                             2.2.6  Flexible bag—Tedlar,1 or equivalent,
                                           with a capacity of 2 to 3 cu. ft. Leak test the
                                           bag in the laboratory before using.
                                             2.2.7  Pitot tube—Type  S, or equivalent,
                                           attached to the probe so that  the sampling
                                           flow rate  can be regulated proportional to
                                           the stack  gas velocity when velocity Is vary-
                                           ing  with  time or  a sample traverse is
                                           conducted.
                                             2.3  Analysis.
                                             2.3.1  Orsat analyzer, or equivalent.
                  PROBE
                                          FLEXIBLE TUBING
                                                                      TO ANALYZER
  LTER(G
FILTER (GLASS WOOL)
                                         SQUEEZE BULB
   3. Procedure.
   3.1  Crab sampling.
   3.1.1  Set up the equipment as shown in
 Figure 3-1, making sure all connections are
 leak-free. Place the probe in the stack at a
 sampling point and purge the sampling line.
   3.1.2  Draw sample into the analyzer.
   3.2  Integrated sampling.
   3.2.1  Evacuate the flexible bag. Set up the
 equipment  as shown In Figure 3-2 with the
 bag disconnected. Place the probe  in the
 stack and purge the sampling line. Connect
 the bag, making sure that all connections are
 tight and that there are no  leaks.
   3.2.2  Sample at a rate proportional to the
 stack velocity.
   3.3  Analysis.
   3.3.1  Determine the  CO., o,. and CO con-
 centrations as soon as possible. Make as many
 passes as are necessary to give constant read-
 ings. If more than ten passes are necessary.
 replace the  absorbing solution.
   3.3.2  For grab  sampling, repeat the sam-
 pling and analysis  until three  consecutive
samples vary no more  than  0.5 percent  by
volume for  each component being analyzed.
   3.3.3  For Integrated  sampling, repeat the
analysis of  the sample  until  three consecu-
tive analyses vary no more than 0.2 percent
by volume for  each component   beinir
analyzed.
  4, Calculations,
  4.1  Carbon dioxide. Average the three con-
secutive runs and report the result  to the
nearest 0.1% r
   4.2  Excess air. Use Equation 3-1 to calcu-
late excess air.  and average the runs. Report
the result to the nearest 0.1% excess air.

 % EA =
                                                                                       0.264(%
                                                                                                                       CO)
                                                                                                                            X100
                        Figure 3-1.  Grab-sampling train.
                                            RATE)
         AIR-COOLED CONDENSER
    PROBE
FILTER [GLASS WOOL)
                                                                  QUICK DISCONNECT
                                  RIGID CONTAINER
               Figure 3-2. Integrated gas • sampling train.
                                                                                                                    equation .'!- 1
                                                                                       where:
                                                                                         %EA=Percent excess air.
                                                                                          %O3:= Percent oxygen by volume, dry basis.
                                                                                          %Na=Percent nitrogen by  volume,  dry
                                                                                                  basis.
                                                                                         %CO=Percent carbon  monoxide  by vol-
                                                                                                  ume, dry basis.
                                                                                         0.264= Ratio of oxygen to  nitrogen in air
                                                                                                  by volume.
                                                                                         4.3  Dry molecular weight. Use Equation
                                                                                       3-2 to calculate dry molecular weight and
                                                                                       average the runs.  Report the result to the
                                                                                       nearest tenth.
                                                                                       Md = 0.44(%CO..)+0.32(%O.)
                                                                                                               +b.28(
-------
                6. Reference*.
                AHshuller, A. P., et a.],, Storage of Oases
              and Vapon In Plastic Bags, Int. J. Air ft
              Water Pollution, 6:75-81. 1983.
                Conner, William D., and J. S.  Nader, Air
              Sampling with Plastlo Bags, Journal of the
              American  Industrial  Hygiene  Association,
              25:291-297,  May-June 1964.
                Devorkln, Howard,  et al., Air  Pollution
              Source Testing Manual, Air Pollution Con-
              trol District, Los Angeles,  Calif., November
              1083.
to
O
                                                                                                               '///////////////////A
§

I
m
O

O
5
                                                                FEDERAL  REGISTER, VOL.  36, NO. 247—THURSDAY, DECEMBER 33, 1971
$
-3

-------
                 APPENDIX B




MOISTURE FRACTION DETERMINATION FROM METHOD 5
                      21

-------
                                           RULES  AND  REGULATIONS
                                                                                                    24889
 ////A
s/777.
'7/77.
'7/77,
/777s
////A
44444
'7/7.
 7-7-,
444
                    ,44
                         7777,
                                 ///y
                                 '777.
                                 77.
                                         777.
      444^444^244
              444
              444
                                444
                                         444
                    777
                                               777,
                                               7ZZ;
                   s///
                                       777
                    777
                                             777;
777>
                                                     '77.'
                                                     7/7;
                          7777
                                                     7777
                                                              77
                                                            4444
                                                            /777
                                                            S/7/
                                                                    777
                                                           777
                                                           777;
                                        777:
             7777
                                                                    777.
              777
              777;
              Y//,
                                                                   777y
                                       Container No. 3. Weigh the spent silica gel
                                     and report to the nearest gram.
                                       5. Calibration.
                                       Use methods  and equipment which have
                                     been  approved  by the Administrator  to
                                     calibrate the orifice meter,  pltot  tube, dry
                                     gas meter,  and probe heater. Recalibrate
                                     after each test series.
                                       6. Calculations.
                                       6.1  Average dry gas meter temperature
                                     and average orifice pressure drop. See data
                                     sheet  (Figure 5-2).
                                       6.2  Dry  gas volume. Correct the  sample
                                     volume measured by the dry gas meter  to
                                     standard conditions (70° P., 29.92 Inches Hg)
                                     by using Equation 5-1.
                                                mi
                                                17-71
                                             in. H6
                                                                   T.
                                                                                 Tm= Average dry gas meter temperature,
                                                                                       "R.
                                                                                Pb.r= Barometric pressure at the orifice
                                                                                       meter, inches Hg.
                                                                                 AH = Average  pressure  drop across  the
                                                                                       orifice  meter, Inches H,O.
                                                                                13.6= Specific gravity of mercury.
                                                                                P. u= Absolute pressure  at standard con-
                                                                                       ditions, 28.92 Inches Hg.

                                                                               8.3  Volume of water vapor.

                                                                                               RT..
                      0.0474

                          equation ~>-2
where:
  V».td= Volume of water vapor In the gas
          sample   (standard  conditions) ,
          cu. ft.
    Vi.= Total volume of liquid collected in
          implngers and silica gel (see Fig-
          ure 9-3). ml.
   on.,o= Density of water, 1 g./ml.
  Mn.,o= Molecular weight of water. 18 lb./
          Ib.-mole.
     B= Ideal  gas  constant, 21.83  Inches
          Hg— cu. ft./lb.-mole-°R.
   T,I4= Absolute temperature at standard
          conditions, 630* R.
   P.tJ=AbsQlute pressure at standard con-
          ditions. 29.92 Inches Hg.

  6.4  Moisture content.
                                                                               B.0=
                                                                                       '-.id
                                                                                    V..U+V...,

                                                                                              equation 5-3

                                                                    where:
                                                                     Bwo = Proportion \>y volume of water vapor in tin-pas
                                                                            stream, dime-iisionless.
                                                                     v*.ui=VoIume of water in the pas sample (standard
                                                                            conditions), cu. ft.
                                                                     v».w=Volume of gas sample tlirougli the dry ciu innter
                                                                            (standard conditians), cu. ft.

                                                                    X*
                                                              . equation &-1
                                     where:
                                       V«,,4= Volume of gas sample through the
                                               dry gas meter (standard condi-
                                               tions) , cu. ft.
                                         V_= Volume of gas sample through the
                                               dry  gas meter  (meter  condi-
                                               tions) , cu. ft.
                                        T.tl= Absolute temperature at standard
                                               conditions, 530* R.
                                 tfOUTO, VOL M, NO. 247—THURSDAY, OECEMUI 23,  1971
                                                       23

-------
24890
   RULES AND  REGULATIONS
                      PLANT
                      DATE.
                      RUN NO,
  CONTAINER
   NUMBER
    TOTAL
                          WEIGHT OF PARTICULAR COLLECTED,
                                        mg
                   FINAL WEIGHT
TARE WEIGHT
                 Figure5-3. Analytical data.

FINAL
INITIAL
LIQUID COLLECTED
TOTAL VOLUME COLLECTED
VOLUME OF LIQUID
WATER COLLECTED
IMPINGER
VOLUME.
ml

-


SILICA GEL
WEIGHT.
9



8TJ nil
 CONVERT WEIGHT OF WATER TO VOLUME BY DIVIDING TOTAL WEIGHT
 INCREASE BY DENSITY OF WATER. (1 g. ml):
                         FEDERAL REGISTER, VOL 36, NO. 247—THURSDAY, DECEMBER 23, 1971
                                                   24

-------
    APPENDIX C




STATISTICAL METHODS
          25

-------
                            APPENDIX C. STATISTICAL METHODS
     This appendix consists of various sections which contain details of statistical procedures carried out in the analysis
of the data. Reference to these sections has been made at various junctures in the body of the report.  Each section is
an independent analysis pertinent to a particular portion of the report.

C.I  Expected Mean Squares

     In an analysis of variance, the mean square for each factor is computed. The mean squares are variance estimates,
and they are used to determine which factors affect the overall mean level.  However, if the expected values of these
variance estimates or expected mean squares are known, the individual variance components of interest can also be
estimated.

     The  basic design used in this study is a nested design with unequal levels of the various factors. That is, the num-
ber of labs per site and repetitions per lab vary from one site to the next.  The expected mean squares for this design
are not determined in a straightforward manner, but rather they are obtained as a weighted average of the varying
sample sizes.  The F-tests obtained using these expected mean squares are inexact with the exception of the lowest or-
der comparison made with respect to the error term. However, that is the only test of interest in this study.

     The  expected mean square  (EMS) for the labs/sites factor consists of a within-laboratory term, a2, and a multiple
of the laboratory bias term, a\.  What is needed is to determine the multiple, k, for a given set of determinations. The
formula was developed by Anderson and Bancroft*s* and is given by
                                               i  I

where

                            HI: is the number of determinations at site / by lab /
                           fii =
                                    df

                           Hj is the total number of determinations at site i

and

                           of/is the degrees of freedom for labs.

     As an example, we have the following sample sizes for the analysis of the Method 3 data:

                      nt =56                    «2 =63                    «3 = 42

                     itn = 14                   «2i = 16                   «3i = 12
                                                                               = 12
                     «14 = 13                   «24 = 15                   «34 = 11
                                                   27

-------
 and there are 9 degrees of freedom. So
                             = y yV"'/""'/ „ ..2
                            '    i   i       9

                             ^fCH)"'
      Thus, we can say that for the labs/sites factor, the expected mean square is a2 + (13.37)a£, and using this rela-
 tionship we can estimate a£ from the AOV table.

 C.2 Precision Estimates For % CO2

      This section presents the analysis of variance table and develops the variance estimates for the % CO2  determina-
 tion. There were 161 determinations used, which results in a total degrees of freedom of 160 for this analysis.  Of
 this number, 2 are attributed to the site factor. The labs/sites degrees of freedom are  obtained by taking the number
 of labs at a site less one  and summing for all sites, or 3(4 — 1) = 9. The remaining degrees of freedom are attributed to
 the error term, which is calculated from the repetitions within each laboratory at each site. The AOV is summarized
 in Table C-l.
       TABLE C.I ANALYSIS OF VARIANCE FOR % CO2.
                                                                  The F value for labs/sites is the ratio of the mean
                                                            square for labs/sites to the mean square for error.
                                                            This ratio exceeds the critical value of 1 .95, taken
                                                            from a table of the F-distribution at the 5 percent
                                                            level of significance with 9 and 149 degrees of free-
                                                            dom. This implies that there is a laboratory effect,
                                                            or equivalently, that the laboratory bias variance, a\ ,
                                                            is greater than zero.

                                                                  Using the expected mean squares, we can ob-
                                                            tain precision estimates for the within-laboratory and
                                                            laboratory bias components. The expected mean
square of the error term is a2 , the within-laboratory variance. Thus, the estimated within-laboratory variance is the
mean square for error, or

                                               a2 = 2.06.

This gives an estimated within-laboratory standard deviation of
Source
Sites
Labs/Sites
Error
Total
DF
2
9
149
160
SS
1222.82
152.84
306.59
1682.25
MS
611.41
16.98
2.06
F
_*
8.24f
EMS
_*
a* + (13.37)a}
o- L
*Not of interest.
fSignificant value at 5% level.
                                                 = N/2~06

                                                 = 1.44%C02 by volume.

This estimate has 149 degrees of freedom associated with it.
                                                    28

-------
      The laboratory bias variance is estimated from the labs/sites mean square.  The manner of obtaining the expected
 mean square is discussed in Appendix C. 1. Since the expected mean square is

                                         EMS = o2 + 13.37a2,

 a\ is estimated by

                                             >2 _MSL-d*
                                             °L     13.37

 where MS/,  is the mean square for labs/sites. Then

                                           f2_ 16.98-2.06
                                                    13.37

                                                14.92
                                                13.37

                                              = 1.12.

 The estimated laboratory bias variance has 9 degrees of freedom associated with it.  The laboratory bias standard devia-
 tion, OL , is estimated as
                                                = 1.06%C02 by volume.

      The between-laboratory variance, a2,, is defined as

                                            °l = oi + o\

Substituting the estimates for a2 and a£ gives

                                             dl = a2 + a2

                                                = (1.12) + (2.06)

                                                = 3.18

and the between-laboratory standard deviation estimate is
                                                = 1.78%CO2 by volume.

C.3  Precision Estimates For % O2

     The analysis of variance table for % 02 determination is presented here, and the variance component estimates
derived. There are 161 determinations used in the analysis, for a total of 160 degrees of freedom. Of these, 3 — 1 = 2
are due to sites, while 3(4 — 1) = 9 are due to labs/sites. The  remainder form the error term. The analysis of variance
table is shown in Table C.2.
                                                    29

-------
        TABLE C.2 ANALYSIS OF VARIANCE FOR % O2.
Source
Sites
Labs/Sites
Error
Total
DF
2
9
149
160
SS
2127.80
225.37
432.55
2785.72
MS
1063.90
25.04
2.90
F
_*
8.63f
EMS
_*
a2 + (13.37)4
a2
*Not of interest.
tSignificant value at 5% level.
                  The F-value shown for labs/sites is the ratio of
            the mean squares for labs/sites and error. The value
            exceeds the tabled value of 1.95 taken from a table
            of the /""-distribution at the 5 percent significance
            level with 9 and 149 degrees of freedom. This implies
            that there is a significant laboratory effect on the O2
            determination,  or that a\ is greater than zero.

                  Using the expected mean squares, the within-
            laboratory and laboratory bias variances may be
            estimated. The expected mean square for error is a2,
            the within-laboratory variance. Thus, the estimated
            within-laboratory variance is
                                              a2 = 2.90,
the mean square for error. There are 149 degrees of freedom associated with this variance estimate. The within-
laboratory standard deviation, a, is estimated as
                                                = 1.70%O2 by volume.

      The expected mean square for labs/sites is a2 + (13.37)a£.  Then o£ is estimated by
                                            8?=-
where MSL is the mean square for labs/sites. Thus,
    13.37


25.04 - 2.90
    13.37

22.14
                                               13.37

                                             = 1.66,

with 9 degrees of freedom. The estimated laboratory bias standard deviation, OL , is
                                                 = 1.29%02 by volume.

     The between-laboratory precision components are estimated from the above. The between-laboratory variance,
  , is estimated by
                                            = (2.90) + (1.66)

                                            = 4.56
                                                    30

-------
The between-laboratory standard deviation, then, is estimated as
                                                = 2.14%O2 by volume.

C.4  Precision Estimates For Dry Molecular Weight.

      The dry molecular weight precision components are estimated by using an analysis of variance on the values in
Table 3.  There were 161 total determinations made, which gives 160 total degrees of freedom. The site factor accounts
for 3 — 1 = 2 of these, and the labs/sites accounts for 3(4 — 1) = 9, the number of labs less one at each site, summed
for the three sites.  The remaining 149 degrees of freedom are attributable to the repetitions per lab, or error term.
The analysis of variance table is shown in Table C.3.
             TABLE C.3 ANALYSIS OF VARIANCE
               FOR DRY MOLECULAR WEIGHT
Source
Sites
Labs/Sites
Error
Total
DF
2
9
149
160
SS
14.62
2.33
5.95
22.90
MS
7.31
0.26
0.04
F
_*
6.25f
EMS
_*
a* + (13.37)4
a2
*Not of interest.
• f Significant at 5% level.
      The F-value given for labs/sites is the ratio of
 the mean squares for labs/sites and error.  This value
 may be said to be significant at the 5 percent level
 if it exceeds a tabled value taken from the F-distribu-
 tion with 9 and 149 degrees of freedom. The
 critical value is 1.95, so that the labs/sites factor has
 a significant effect on the overall mean level. This is
 equivalent to saying that the laboratory bias variance,
 0£, is greater than zero.

      To estimate the precision components, the
 expected mean square (EMS} column is used.  The
EMS of the error term is o-2, the within-laboratory
 variance. Thus,
                                                    error
                                               = 0.04

is the estimated within-laboratory variance, with 149 degrees of freedom. The within-laboratory standard deviation,
then, is estimated by
     The EMS of the labs/sites factor is a2 + (13.3 7)o£, as developed in Appendix C.I. Thus, the estimated labora-
tory bias variance is
                                                        2
where MSL is the mean square for labs/sites. Then
                                                MSL - a
                                                  13.37
                                                0.26 - 0.04
                                                   13.37
                                                   31

-------
                                                0.22
                                           07 - -
                                            L  13.37


                                             = 0.02.


The laboratory bias variance estimate has 9 degrees of freedom associated with it. The estimated laboratory bias standard

deviation is
                                                = 0.141b/lb-mole.


      The between-laboratory components are estimated using the above estimates. The between-laboratory variance

is defined as
so that the estimated value,


                                         &* = a*+&2


                                           = (0.02) + (0.04)


                                           = (0.06).


This gives an estimated between-laboratory variance of
                                                = 0.24 Ib/lb-mole.


C.5  Distribution of Excess Air.


      The excess air determinations were used to propose a distribution model for these determinations which could

predict the excess air percentage at a given level of O2.  The formula in the absence of CO, as derived in Section IV

of the report,


                                             %02
                      %EA = - X 100 percent
                             (26.4) - (1 .264) % 02 - (0.264) % CO2


indicates that the chief contribution to the %EA was made by % O2 . A model was proposed, then, that did not con-

tain the % CO2 as an independent variable. Due to the nature of the determinations, an exponential model,
was proposed where


                                             y is%EA


                                             x is % O2


and


                                     a0, a i, are unknown constants.



                                                   32

-------
     A least squares regression^6* is used to estimate a0 and a, . Taking the natural log of the equation gives
which is of the form
This is the usual form of a simple linear regression, and by using the formulas



                                                   E>.    — i
                                                Xjyj/n - xy
and
where



                                    n is the total determinations used




                                    3c is the mean % O2



                                    y' is the mean of the (^j')'s




the least squares estimates of a'0 and Oi  are obtained. These are derived using one quarter of the data points.




                                               a, =0.21




                                               a'0 = 2.35




Then to fit the proposed model,




                                               a'0 =&n.a0




which implies







                                                = e(2.35)




                                                = 10.47.




Thus, we have
     The closeness of this model to the determinations is measured by the coefficient of determination, r2. The

formula for r2 is
                                            n                  ~|

                                            £  (xt-^W-y'n
                                            ='J
                                         1=1          ii = 1




                                                    33

-------
The value of r2 was 0.993, which indicates an extremely good fit to the model.

C.6  Moisture  Fraction Precision Estimates.

      The moisture fractions which appear in Table 5 are used to develop precision estimates. The determinations are
used in an analysis of variance on a nested design, with 150 total determinations used. This gives 149 total degrees of
freedom. Of these, 2 are attributed to sites, while the labs/sites factor accounts for (4 — 1) + (4 — 1) + (3 — 1) = 8
degrees of freedom, the number of labs per site less one, summed over all sites.  The remaining 140 are attributed to
the repetitions per lab or error term.  The analysis of variance is summarized in Table C.4.
            TABLE C.4 ANALYSIS OF VARIANCE
                FOR MOISTURE FRACTION
Source
Sites
Labs/Sites
Error
Total
DF
2
8
140
150
SS
1.699
0.103
0.136
1.938
MS
0.8SO
0.013
0.001
F
_*
13.000t
EMS
_*
o3 + (13.78)ai
o1
•Not of interest.
fSignificant at 5% level.
     The F-value shown is the ratio of the labs/sites
mean square to the error mean square.  This value is
significant at the 5 percent level if it exceeds the
tabled value taken from the /"-distribution with 8
and 140 degrees of freedom.  The table value is
approximately 2.02, which implies that there is a
significant laboratory effect, or that the laboratory
bias variance, 0£, is greater than zero.

     Using the expected mean squares, the variance
components may  be estimated. The EMS of the error
term is a2, the within-laboratory variance. Thus the
estimated value of a2 is
                                              a2 = 0.001
the error mean square. This estimate has 140 degrees of freedom associated with it. Then the estimated within-
laboratory standard deviation is
                                                = 0.032.

     The EMS of the labs/sites factor is a2 + 13.78o|. The factor 13.78 is obtained according to the formula in
Appendix C.I, substituting the values for the sample sizes and degrees of freedom for this study. Thus

                                         E(MSL) = a2 + 13.7802,

where MSi is the mean square for labs/sites. This implies

                                         5^MSL-o2
                                          1    13.78

                                              0.013-0.001
                                                 13.78
                                              0.012
                                              13.78

                                             '0.001
                                                   34

-------
with 8 degrees of freedom. The estimated laboratory bias standard deviation, then, is




                                             oi=V0.001




                                                = 0.032.




      The between-laboratory variance is o2b = o2L 4- a2 .  Substituting into this equation, gives




                                         dl=ai + o*




                                            = (0.001) + (0.001)




                                            = (0.002).




From this, the estimated between-laboratory standard deviation is
                                                = 0.045.
                                                    35

-------
                                     LIST OF REFERENCES
1.    Environmental Protection Agency "Standards of Performance for New Stationary Sources," Federal Register,
     Vol. 36, No. 247, December 23, 1971, pp 24876-24893.

2.    Harnil, H.F. and Camann, D.E., "Collaborative Study of Method for the Determination of Particulate Matter
     Emissions from Stationary Sources (Portland Cement Plants)," Southwest Research Institute report for
     Environmental Protection Agency, in preparation.

3.    Hamil, H.F. and Thomas, R.E., "Collaborative Study of Method for the Determination of Particulate Matter
     Emissions from Stationary Sources (Fossil Fuel-Fired Steam Generators)," Southwest Research Institute
     report for Environmental Protection Agency,  June 30,1974.

4.    Hamil, H.F. and Thomas, R.E., "Collaborative Study of Method for the Determination of Particulate Matter
     Emissions from Stationary Sources (Municipal Incinerators)." Southwest Research Institute report for
     Environmental Protection Agency, July  1, 1974.

5.    Anderson, R.L. and Bancroft, T.A., Statistical Theory in Research. McGraw-Hill, New York, 1952.

6.    Dixon, W.J. and Massey, F.J. Jr., Introduction to Statistical Analysis, 3rd Edition.  McGraw-Hill, New York,
     1969.

7.    Mitchell, WJ. and Midgett, M.R., "Studies of the Field Reliability of the Orsat Analyzer," Environmental
     Protection Agency, QAEML, (To be published).
                                                   37

-------
                                             TECHNICAL REPORT DATA
                                    (Please read Instructions an the reverse before completing)
1. REPORT NO.
    EPA-650/4-73-026
                                                                            3. RECIPIENT'S ACCESSION"NO.
4. TITLE AND SUBTf'.E
    Collaborative Study of Method for Stack
    Gas Analysis and Determination of Moisture Fraction
    with Use of Method 5
                 5. REPORT DATE
                    June 1974
                 6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
    Henry F. Harnil
    Richard E. Thomas
                 8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
    Southwest Research Institute
    8500 Culebra Road
    San Antonio, Texas  78284
                 10. PROGRAM ELEMENT NO.

                    Task Order 8
                 11. CONTRACT/GRANT NO.

                    68-02-0626
 12. SPONSORING AGENCY NAME AND ADDRESS
    Quality Assurance and Environmental Monitoring Laboratory
    National Environmental Research Center
    Research Triangle Park, North Carolina  27711
                 13. TYPE OF REPORT AND PERIOD COVERED
                    Task Order
                 14. SPONSORING AGENCY CODE
 15. SUPPLEMENTARY NOTES
 16. ABSTRACT
          Statistical analyses are performed on data from EPA Method 3 (Stack Gas Analysis for Carbon Dioxide, Excess Air
    and Dry Molecular Weight) and from the stack gas moisture fraction determination obtained in the collaborative testing
    of EPA Method 5 (Particulates).  Using data from Method 5 tests at a Portland cement plant, a coal-fired power plant
    and a municipal incinerator, estimation is made of the precision that can be expected with the use of these methods. For
    Method 3, the precision of CO2 and 02 determination using an Orsat analyzer is investigated, as well as the effect of
    this on the dry molecular weight and excess air calculations. In addition, the effect of variability in CO2 and 02 deter-
    minations on correcting particulate determinations to a common base is studied. The precision of the determination of
    the moisture fraction of the stack gas by the formula in Method 5 is studied.  Recommendations are made for the improve-
    ment of the precision of the Orsat method.
17.
                                         KEY WORDS AND DOCUMENT ANALYSIS
                       DESCRIPTORS
                                                           b.IDENTIFIERS/OPEN ENDED TERMS
                                   c.  COSATI Field/Group
    Air Pollution, 1302
    Flue Gases
    Collaborative Testing
    Methods Standardization
    Orsat
13-B
07-B
18. DISTRIBUTION STATEMENT
     Release Unlimited
                                                           19. SECURITY CLASS (This Report)
                                                                  UNCLASSIFIED
                                   21. NO. OF PAGES

                                          45
20. SECURITY CLASS (This page)
       UNCLASSIFIED
                                                                                              22. PRICE
EPA Form 2220-1 (9-73)

-------