EPA-650/4-74-039
OCTOBER 1973
Environmental Monitoring Series

-------
                                EPA-650/4-74-039
           LABORATORY
   AND  FIELD  EVALUATIONS
                   OF
 EPA  METHODS 2,  6,  AND  7
                    by

               Henry F. Hamil

           Southwest Research Institute
              8500 Culebra Road
            San Antonio, Texas 78284
            Contract No. 68-02-0626
          Task Order 1, Change Order 1
               ROAPNo. 26AAG
           Program Element No. 1HA327
        EPA Project Officer: M.R.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

-------
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.
                                11

-------
                           ABSTRACT









       A study was made to evaluate Methods 2,  6, and 7,  proposed by




the Environmental Protection Agency for determination of stack gas




velocity  and volumetric flow rate,  sulfur dioxide emissions, and nitrogen




oxide emissions,  respectively.  These evaluations were conducted prior




to collaborative testing of the subject methods.  Findings and conclusions




concerning these methods are  given below.







Method 2 - Stack Gas Velocity and Volumetric Flow Rate




       Statistical analysis  of stack gas velocity data  indicates that




Method 2 provides an accurate estimate of the true stack gas velocity at




high gas velocities.  Accuracy of Method 2 velocity estimates at low gas




velocities is shown to be  unreliable.  Correlation analysis demonstrates




that the volumetric flow rate estimates have the same characteristics as




the velocity estimates.  Correlation analysis also demonstrates that the




variation in the stack gas velocity and volumetric flow rate estimates is




principally due to variation in determination of Ap.the velocity head in




the stack.  By way of comparison, a separate analysis was performed on




individual velocity traverse data.





Method 6 - Sulfur Dioxide




       Investigation of possible  causes of variation in collection efficiency




of SO, in Method 6 was  made.  Com entration of SO, in the stack gas is
                              ill

-------
 shown to be the only factor to have any significant effect on collection




 efficiency.  The purge period specified in Method 6 was  shown to be




 necessary to avoid apparent low SC^ values due to retention in the




 isopropanol bubbler. The minimum detectable limit is estimated to be




 3ppm.







 Method 7  - Nitrogen Oxides




       Investigation of possible interference with NOX determination by




 chloride ion indicated the degree of interference to be linearly related




to chloride ion concentration.




       The minimum detectable limit for Method 7 is estimated to  be




 2ppm NOX as NC>2.  The upper limit without dilution is  approximately




 100 ppm NOX as NC>2. The maximum sensitivity  of Method 7 can




approach 0.2 ppm   but probably lies between 0.2 ppm and 2.0 ppm NOX  as




N02.
                               iv

-------
                     TABLE OF CONTENTS



                                                    Page

I.      INTRODUCTION                                1

II.     RESULTS AND CONCLUSIONS                   3

       Method 2 - Stack Gas Velocity and
       Volumetric Flow Rate                           3

           1.  Evaluation of Method 2                   3
           2.  Analysis of Single Traverse Data        23
           3.  Conclusions                            30

       Method 6 - Sulfur Dioxide                       31

       Method 7 - Nitrogen Oxides                     41



APPENDICES I, H,  III and IV

-------
                       LIST OF TABLES
Table                                                        Page

  1.   Operating Characteristics, T.H.  Wharton Power           5
       Plant,  Unit No. 1, Houston Lighting & Power Co.

  2.   Velocity Traverse Data, Houston  Lighting & Power        6
       Co., T. H. Wharton Power Plant, Unit No.  1, Low
       Fuel Feed Rate

  3.   Velocity Traverse Data, Houston  Lighting & Power        7
       Co., T. H. Wharton Power Plant, Unit No.  1,
       High Fuel Feed Rate

  4.   Experimental and Theoretical Gas Velocity and           11
       Volumetric Flow Rate - Houston Lighting and Power
       Company, T. H. Wharton  Plant, Unit No. 1

  5.   Correlation Coefficients Vs as a Dependent Variable      12
       of Various Experimental Parameters

  6.   Statistical Analysis - Experimental Stack Velocities
       and Related Parameters                                 15

  7.   Statistical Analysis,  Theoretical Stack Velocities and
       Related Parameters                                     21

  8.   Velocity Traverse Data, Houston  Lighting & Power Co.,
       T. H. Wharton Power Plant, Unit No. 1,  Low Fuel       24
       Feed Rate, Single Diameter Traverse

  9.   Velocity Traverse Data, Houston  Lighting & Power Co.,
       T.H. Wharton Power Plant, Unit No. 1, High Fuel       25
       Feed Rate, Single Diameter Traverse

 10.   Experimental and Theoretical Gas Velocity and
       Volumetric Flow Rate - Houston Lighting and Power       26
       Company,  T.  H. Wharton  Plant, Unit No. 1,  Single
       Diameter Traverse

 11.   Statistical Analysis,  Experimental and Theoretical
       Stack Velocities and Related Parameters, Single          27
       Diameter Traverse
                              vi

-------
                      List of Tables (cont'd.)
                                                             Pae
12.   Fractional- Factorial Experiment Design for Five
      Variables in Eight Experiments Showing the Values
      for Each Independent Variable and the Dependent
      Variable                                                 33

13.   Intermediate Experimental Data from Fractional-
      Factorial Experiment Design                              36

14.   Statistical Analysis of Fractional-Factorial Design         38

15.   Data Tabulation—Interference of Hydrogen Chloride
      with NOX  Determination - EPA Method No. 7              42
                               vii

-------
                       I.  INTRODUCTION







        This report describes the work performed and the results




obtained on Task Order No. 1, and Task Order No. 1, Change Order




No.  1,  which included evaluation of the methods for the determination




of stack gas velocity and  volumetric flow rate,  the determination of




nitrogen oxide  emissions, and the determination of sulfur oxide  emissions




in fossil-fuel fired steam generators (Federal Register, December 23,




1971).




        Complete plans were developed for the accomplishment of the




objectives before the experimental work began.  These plans were




submitted to the Project Officer  by letter  dated October 3, 1972, and




received subsequent approval.




        The task order  required  experimental investigation of the




following: Possible sources of error in determination of stack  gas




velocity and volumetric flow rate due to calibration of the type S pitot




tube, and determination of precision and accuracy of  the method in a




suitable facility in which  a theoretical  value for velocity  could be




obtained.





        The laboratory  investigation of the sulfur dioxide method included




an investigation of SO£ collection efficiency as a function of changing




concentration,  as well  as investigation of low recovery of SOj> due  to




retention in the isopropanol bubbler.

-------
        The laboratory investigation of the nitrogen oxide method




included an evaluation of the detection limits of the method and an





investigation of possible chloride ion interference in the analysis.





       A glossary of appropriate equations and terms used in this




report  is  given in Appendix I.

-------
                II.  RESULTS AND CONCLUSIONS







METHOD 2-STACK GAS VELOCITY AND VOLUMETRIC FLOW RATE




1.     Evaluation of Method 2




       The stack gas velocity and volumetric flow rate as determined by




Method 2 are used in conjunction with the methods for determination of




specific  pollutants to determine the emission rate of those pollutants.




Therefore, evaluation of Method 2  with regard to precision and accuracy




was   considered necessary.  The experimental program was structured




to allow  estimation of the precision and accuracy of Method 2.




Statistical  analysis of the data obtained in the experimental program was




performed in order to determine which experimental variables contributed




most to the variation in the stack gas velocity.




       Permission was obtained from Houston Lighting and Power Company




to perform a series of measurements at their T. H. Wharton Power Plant.




This plant  is a natural gas fired steam generating plant, normally maintained




on hot standby,  and is used to balance peak loads. As a result,




arrangements could be made  to make velocity traverse measurements at




peak load and at a lower level,  to give two different stack velocities during




the evaluation.




       The fuel gas feed rate is accurately measured at the T. H. Wharton




plant, and this value, in conjunction with stack gas composition as




determined by Or sat analysis, allows the calculation of theoretical stack

-------
gas velocities for comparison with the experimentally determined




values.  A sample calculation demonstrating the method used to




determine the theoretical stack gas velocity and volumetric flow rate




is shown in Appendix II.




       Table 1 tabulates the pertinent operating characteristics of the




power plant unit which was used for field investigations, while Tables 2




and 3 show the data obtained by pitot tube traverses at this facility, at




the two feed  rates studied.

-------
    TABLE 1.   OPERATING CHARACTERISTICS
   T. H. WHARTON POWER PLANT, UNIT NO.  1
     HOUSTON LIGHTING & POWER COMPANY
Output:                 75 megawatts

Fuel Consumption:      13, 600 cu ft per minute, natural gas

Air Rate:               146,000 cu  ft per minute

Steam Rate:             650,000 Ib per hour, 1340 psi,  955°F

Stack Velocity:          52 ft per second

Sample Ports:           Two 3-in.  ports at 90-degree spacing

      are located next to a walkway handrail  at the  100-ft level

      (above grade).  This location is 56 feet (8. 3 diameters)

      above the  preheaters  for the boiler,  Two 45-degree elbows

      in the vertical stack run affect the flow pattern somewhat,

      but the ports are still 24 feet (3. 6 diameters) above the

      higher ell so that a reasonably uniform flow pattern should

      exist at the existing sample port elevation.

-------
                                     TABLE 2.  VELOCITY TRAVERSE DATA,
                       HOUSTON LIGHTING & POWER CO. ,  T. H. WHARTON POWER PLANT,
                                       UNIT NO. 1, LOW FUEL FEED RATE1
Run °v& av*,, s xs *s vs
Date No. Time (in H2O) (in H2O)Z OF OF in H^ %H2O % COZ %O2 Md Mg ^g S(
5/22/73 1 1125-1155 0.275 0.524 266 726 29.83 14.0 8.2 6.4 29.5 27.9 32.2
2 1210-1235 0.285 0.521 252 712 29.82 14.6 8.5 5.0 29.5 27.8 31.9
5/23/73 5 1025-1047 0.245 0.498 255 715 29.78 13.5 7.8 6.5 29.5 28.0 33.6
" 6 1055-1130 0.265 0.513 251 711 29.77 14.0 8.2 6.3 29.5 27.9 31.8
" 7 1135-1200 0.265 0.517 264 724 29.77 14.0 8.2 6.4 29.5 27.9 32.4
Q
:f/hr x 10~<
4.970
4. 937
4.774
4.914
4.911
1
 Definition of symbols is given in Appendix I.

-------
                                    TABLES.  VELOCITY TRAVERSE DATA,
                       HOUSTON LIGHTING k POWER CO. ,  T.  H. WHARTON POWER PLANT,
                                      UNIT NO. 1,  LOW FUEL FEED RATE1
                       Apavg   **Vavg    Ts   Ts    p

Date TraNTSe   Time     ^ ^  °F   °R   in Hg    %H2
                                                                     %CQ
5/2Z/73   3    1400-1425   0.88     0.912    307   767   29.80    14.9     8.8

          4    1428-1455   0.82     0.878    317   777   29.80    14.3     8.4

5/23/73   8    1340-1400   0.89     0.937    300   760   29.73    14.7     8.6

          9    1405-1425   0.89     0.937    310   770   29.73    17.1     8.9
                                                                                                V
%0
            M
5.5  29.6   27.9  57.6

5.8  29.6   27.9  55.8

4.9  29.5   27.8  59.2

5.4  29.6   27.6  59.8
                                                                                                          QS
                                                                                                .         , .     
-------
        The sample ports, designated east (E) and west (W) are located

 90  apart on both stacks A and B (see Figure 1).  The total gas flow from

 the unit is split at the preheater outlet and vented through the two stacks
             Jj;
 of 80-in. I. D.  Since there is  a disturbance approximately 3. 6 diameters

 upstream of the sample ports,  24 traverse points were chosen on each

 diameter.

        The Type S pitot tube was calibrated in  a wind tunnel at

 Southwest Research Institute.   Calibration was performed over a velocity

 range of 17 to 70 fps.   At velocities in that range, the pitot tube coefficient

 C   had an  average value of 0. 77.  Variation of the coefficient over the
  P
 working range was within the + 5% specified in  the method.   The coefficient

 was determined with each leg of the pitot tube facing  the gas flow and

 was found to be the same in  each case.

        Data were taken on two days, at two different generating levels

 each day.  The higher generating level of 71.0  megawatts was near the

 rated peak generating  level of  72 megawatts, while the lower level was

 40.0  or 41.  5 megawatts. On the second day, two traverses were made

 during unit line-out while the generating levels  were  39.0 and 42. 0 megawatts.

 The remaining lower level generating loads were 40.0 megawatts after

 line-out.

       Fuel feed rates were 680 MCFH at the 71 megawatt generating level

 and 385-390 MCFH at  the 40-41. 5 megawatt generating level.
 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 these factors ure
located in Appendix 1.

-------
FIGURE 1.   T. H.  WHARTON PLANT, UNIT NO.  1,
        SAMPLE PORT CONFIGURATION

-------
                                                                 10
        Integrated stack gas samples were taken on each traverse by a




 modification of Method 3 in which a squeeze bulb was utilized to




 transfer gas from the stack into the gas sample bag. A gas sample tube




 was incorporated into the probe bundle, which consisted of the pitot tube




 thermocouple, and gas sample tube, with the inlet position near the tip




 of the Type S pitot tube.  Equal volumes of gas were withdrawn via the




 squeeze bulb at each traverse point to provide an integrated gas sample




 which was analyzed by Or sat analysis.




        The data  obtained in the experimental runs are presented in




 Tables 2 and 3, along with the experimental values of stack velocity and




 volumetric flow rate. Stack velocity and volumetric flow rate were




 calculated in accordance with the Federal Register,and the appropriate




 equations are given in Appendix I.




        In Table 4 are presented the theoretical values of stack velocity




and volumetric flow rate which were calculated using fuel  feed rate,




 stack gas composition, fuel composition, stack gas temperature and




stack dimensions.  Also presented in Table 4 are the experimental values




for stack velocity and volumetric flow rate to allow a visual comparison.




        In an attempt to determine which experimental parameters have




the greatest influence on the values for V , a correlation analysis using




the data in Tables 2 and 3 was performed using V  as the dependent
                                               s



variable and Ap   ,  ( 'Y A P)avg«  Ts.  Pg,  and Mg  as independent




variables.

-------
                    TABLE 4 .   EXPERIMENTAL AND THEORETICAL GAS VELOCITY
                                      AND VOLUMETRIC FLOW RATE -
                               HOUSTON LIGHTING AND POWER COMPANY,
                                  T. H.  WHARTON PLANT, UNIT NO. 1
Run
No.1
1
2
3
4
5
6
7
8
9
Stack
Designation
A
B
B
A
A
B
A
A
B
Fuel Gas3
Feed Rate
MCFH
390
390
680
680
385
385
385
680
680
Load,
megawatts
41.5
41. 5
71.0
71.0
40. 5
40.0
40. 0
71.0
71.0
V
• Stack Gas Velocity, fps
Experimental
32.2
31. 6
57. 6
55.8
30.3
31.2
31.8
59.2
59.8
4
Theoretical
32.9
31.9
56. 5
59.8
33.6
31.8
32.4
57. 3
56.1
Q
Volumetric Flow
Experimental
4. 970
4. 937
8. 318
8.021
4. 774
4.914
4.911
8. 683
8. 370
Rate, scf/hr x 10
Theoretical5
5.197
5. 134
8.443
8.975
5. 393
5. 130
5. 130
8. 639
8. 348
*Run data are calculated from appropriate diameter traverse data,  i. e. , Run 1 from traverses 2AW + 1AE.
 Stack designation per Figure 1.
3Fuel gas volume is in cubic feet at 60°F and 1 atmosphere pressure.
^Calculated by the procedure shown in Appendix 1, wet gas basis.
^Calculated by the procedure shown in Appendix 1, dry gas basis.

-------
                                                                 12
        The correlation coefficients obtained are presented in




Table 5.
        TABLE 5.  CORRELATION COEFFICIENTS Vs  AS A


              DEPENDENT VARIABLE OF VARIOUS

                 EXPERIMENTAL PARAMETERS
       Independent Variable,         Correlation Coefficient,


              X	               Vs = *(*)	




             APavg                      0.9944







           (i~Ap^vg                      0.9996





             Ts, °R                      0.9286





             Ps , in.  Hg                 -0.4778





                Ms                      -0.3780







       As can be seen from  the values in Table 5, the strongest




correlation is obtained with V   as a function of (VApL.._   and of
                            S                      a. V g



APavg .   This  correlation analysis shows a very strong linear




relationship between Vg and these two related experimental parameters.




The direct relationship of Vg to the other experimental parameters is




less strong, based upon the correlation coefficients. From the correlation




analysis,  one would conclude that the variance in Vg would be most




affected by the variance in (V A Pj   » which relates directly back to the
variance in Ap&vg.

-------
                                                                 13
       To further check this hypothesis, a simple statistical analysis



of the experimental  stack velocity and volumetric flow rate along with



  Pavg i   ( V  ^ P^vg'    save'  ^s« anc* PS was *nade in which the mean,



variance,  standard deviation, and percent distribution about the mean at



2s were  calculated.  The analysis was  performed on the run data



presented in Tables 2 and 3.
       Percentage distribution about the mean at 2s (i.e.,  the 95%
'0
                               2s •  100
confidence level) is defined as 	  .   It measures the amount of

                                  x

variation in the experimental data for variable x, expressed as a



percentage of the mean value. Each of the independent experimental



parameters being studied is directly or inversely proportional to Vs  in the



Vs  equation.  Thus, a valid technique for determining the parameters to



which Vs is most sensitive, and to which its variability and its uncertainty



are most closely related, is to compare the percentage distribution about



the mean for Vs to the  percentage distributions  about the mean for the



various independent experimental variables.  For comparison, a similar  >



analysis of the theoretical stack velocity was made  along  with the mole



percent CO2 in the stack gas and the stack gas temperature.  These two



parameters were chosen for the latter analysis inasmuch as they are the



two experimentally determined numbers which have the most influence on



the calculation of the theoretical velocity (see  Appendix II). The portion



of the analysis pertaining to Qg is included to  show  the  relationship



between Vg  and Qg and between Qg  and the experimentally determined



parameters.

-------
                                                                 14
       For the statistical treatment, the data were divided into high




and low levels corresponding to the power generating levels of 71.0




and 40.0-41.5 megawatts.




       The results of the analysis  are presented in Table 6, showing




the mean,  variance,  standard deviation, and percentage distribution




about the mean at the 95 percent confidence level.

-------
TABLES-   STATISTICAL ANALYSIS -
EXPERIMENTAL STACK VELOCITIES
    AND RELATED PARAMETERS
Fuel Feed
Rate
Variable Mean
x x
Low Vg 31.4
A Qs 4.901xl06
Apavg 0.26
(-/,/V ) 0.51
Tsavg 717
Ps 29.50
Ms 27.9
High Vg 58.1
Qs 8.348x10
APavg °-87
ra^avg 0/92
Ts 768
avg
PS 29'56
Mc 27.8

Variance
s2
0. 57
.0056x 1012
0.0003
0.0001
52.0
0.0
0.005
4.06
0.0735xl012
0.001
0.0004
44.0
0.0023
0.02

Standard
Deviation
s
0. 756
.075 x 106
0.0175
0.01
7.21
0.0
0.07
2.015
0.271 x 106
0.032
0.02
6. 6
0.048
0. 14

Percent Distribution About the Mean
2 s x 100
x
95% Confidence Limits
4.8
3.7
13. 5
3.9
2.0
0.0
0. 5
6.9
6.5
7.4
4.3
1. 8
0. 3
1.0

-------
                                                                  16
         It can be seen that for /\p   ,  the values of s  and s are not
                                   avg

the same at the high and low levels, indicating an inequality of variance


at the levels studied.  The percent distribution about the mean for



Ap    is + 13. 5 at the low level and + 7. 4 at the high level.  The
    avg

percent distribution about the mean is related to the repeatability of


the method of measuring Ap    .   In this particular experiment,  however,
                            avg

the true value of /\p in the stack  is not  known,  and due to normal


variations in process parameters  the true value  of Ap would be expected


to vary with time.  The natural variations in true Ap with time would



contribute to the variance about the mean for the set of runs and as a


result would be incorporated in the percent distribution about  the mean,


along with that variance about  the  mean attributable to the  repeatability


of the method.


         Since  no procedure is available for readily separating the



effects of variation of true Ap with time from the effect of repeatability


of the method in  the statistical analysis, the percent distribution about  the


mean is considered a measure of the total uncertainty of the  experimental



values, and consists of the uncertainty due  to variations in true  /\p plus


uncertainty due to repeatability of the method.   However,  it is believed



that the flow conditions in the stacks at the  test  site are generally


characteristic  of conditions encountered in  stack velocity measurements,



and as a  result,  the data developed in this study are representative of the


results which can be expected from use of Method 2.

-------
                                                                  17
         The percent distribution about the mean for    p    indicates
                                                       avg



that the total uncertainty in  Ap    is  considerably greater at low values
                               avg


of AP     than at high values of Ap   •  This is not unexpected in view
       avg                         avg


of the behavior of the  Type S pitot tube and inclined manometer when




measuring velocity heads in disturbed  flows.  It has been our experience




that when this measuring system is used in stacks where flow  patterns




are disturbed and/or cyclonic,  oscillations are set up in the liquid




column of the inclined manometer.  The observed oscillations were




0. 2-0. 3 inches of water and were unsymmetric within the observed




range.  Readings  of Ap at each traverse point are made by a visual




estimate of the average  value within the  range of oscillation.   The




range of oscillation at a particular site does not appear  to be a function




of overall velocity.  For example, in the study under discussion, the




range of the manometer oscillation was essentially the same at stack




gas velocities of about 31 and 58 fps.




         As a result,  the relative error  in reading Ap is greater at




low values where  the range of oscillation of the manometer is




approximately equal to the value of Ap  than at high values of Ap




where  the range of oscillation is approximately one-half the value of




AP,



         A brief study was made during pitot tube calibrations to




determine if a Magnehelic^'differential pressure gauge, which

-------
                                                                  18
 incorporates small orifices in the pitot line  connections to provide


 damping of sudden transient pressures, was a suitable substitute for the inclined


 manometer.  The flow in the wind tunnel was intentionally disturbed to


 simulate stack gas flow conditions.  Even  though the Magnehelic gauge


 is damped, resulting /\p measurements were no better than those


 obtained with an inclined manometer inasmuch as the Magnehelic


 gauge showed the characteristic oscillations described above.


         Even though the total uncertainty  in Z\p     appears rather
                                                avg

 large at the lower values observed, the effect on the total uncertainty


 of the experimental stack velocity is minimized to  a fair extent by



 the fact that ("/Z\p)    is used in the calculation of the velocity.  The
                     O
calculation of ( YAp)a   from the individual Ap values measured at


each traverse point is a transformation that both stabilizes the data


and  minimizes its random measurement variation.  As can be seen


in Table 6, equality of variance  is not obtained for (V/!\p)avg over


the range studied.  Instead,  s and s  appear to be a function of the


level of (V/Ap)avg.   This is further shown by examination of the


percent distribution about the mean for ( V/\p )    at the high and
                                               O

low levels, where it can be seen that  the total uncertainty is -f 3.9%


at the low level and f 4. 3% at the high level.  When  the experimental


stack velocity data in Table 6 are examined, it can be seen that s2 and s


also are a  function of the level of Vg.   The percent distribution about


the mean velocity at the high and low  levels is + 6. 9 and + 4. 8, respectively.

-------
                                                                 19
 These values are in iair agreement with the similar values for




 ("v  & p)     at the high and low levels, as would be expected from the
        avg



 highly linear relationship of these two variables indicated in the




 correlation analysis .   The percent distribution about the mean for Q  at




 the high and low levels is   +_6.5 and _+ 3.7, respectively.  These values




 correspond favorably with those same values for Vs and (iA p)    .A
                                                               O


 correlation analysis using Qg =  f(V&) gives  a correlation coefficient of




 0.9984, indicating a strong linear relationship between Qs and Vg, which




 establishes the dependence of Qs on those same parameters upon which




 Vs  is dependent.




       The other parameters used to calculate V  are T       M  »

                                               S       S(avg)    8


 and P .  The correlation analysis  showed poorer linear correlation




 between these variables and V  .  This is supported  by the statistical




 analysis of these parameters.   As  can  be seen in Table 6, the percent




 distribution around the mean for these  variables is small, indicating that




 the total uncertainty is small, and  when variations in process parameters




 are considered to occur, this implies good repeatability for the




 determination of T      , p  , and M .   Thus it would appear that the
                  s/    .   s        s
experimental parameter which most affects the value of V  is (V~AP)
                                                       s           avg



       The theoretical stack velocities were calculated as  shown in




Appendix II.  From the calculation method used, it can be seen that the




calculated  theoretical velocities are dependent upon carbon in the fuel,




carbon in the stack gas, fuel feed rate,  stack gas temperature, and stack



dimensions .

-------
                                                                  20
        Fuel carbon analyses and fuel feed rates were provided by




 Houston Lighting and Power Company,  and no estimate of accuracy





 or repeatability for these values is available.  Fuel feed rates were





 measured with a calibrated orifice meter and were reported as cubic




 feet at 60°F and  1 atmosphere pressure.   Carbon in the feed was




 calculated from a gas chromatographic analysis of the fuel gas and




 was reported as Ib-atoms carbon per Ib-mole fuel gas.




        Carbon in the stack gas was obtained from the Orsat analysis




 of the integrated gas sample  taken on each run.  Stack temperature




 was also determined experimentally on each run.  These two variables





 were subjected to statistical  analysis along with the theoretical





 velocities.





       In Table 7, it can be seen that the percent distribution about




the mean for the theoretical velocity at the high and low levels is jf 5.8




and+_ 4.6,  respectively.  The percent distribution about the mean for the




theoretical volumetric flow rate at the high and low levels is + 6.4 and




+ 4.4, respectively.   The percent distribution about the  mean for CO,
•v.                                                                  £




concentration at the high and low levels  is +_ 5,0 and +  6.1, respectively.





The distribution about the mean for Tg at the high and  low levels is




•f 1.8 and +2.0 percent, respectively.   These values would appear to





indicate that the experimental variable with greatest effect on the




calculated values of theoretical velocities  is the CC»2 content of the  stack





gas as determined by Orsat Analysis and that the accuracy of this





experimental value is reflected in the accuracy of the theoretical velocities

-------
                                  TABLE 7.   STATISTICAL ANALYSIS,
                                 THEORETICAL STACK VELOCITIES AND
                                         RELATED PARAMETERS
                                                                     Percent Distribution About the Mean
Standard
Fuel Feed
Rate
Low


Variable
x
V
Q
%CO0
Mean
x
32.
5.
8.
5
197xl06
2
Variance
0.
0.
0.
56
013xl012
062
Deviation
s
0.
0.
0.
75
113 x 106
25
L s
x 1UO
x
95% Confidence
4.
4.
6.
6
4
1

Level



                             717
                      52.0
                        7.21
                                                                                 2.0
High
V
57.4
2.75
                                                    1.66
                                                                5. 8
%co2
Q
T_
8.7

8.601x10
768

0.
0.

05
077 xlO12
44
0.
0.
6.
22
277 x
6

106

5.
6.
1.
0
4
8

-------
                                                                  22
        Statistical analysis performed on the volumetric flow rate




 data indicates they are primarily dependent on the stack gas velocity.




 The accuracy, repeatability and distribution about the mean for




 volumetric flow rates appear to be a direct function of the same




 parameters as for the velocity data.




        In order to assess the accuracy of Method 2 determination of




 stack gas velocity, a comparison was made between the experimental




 and theoretical stack gas mean values at both the high and low levels.




 The details of the test are shown in Appendix 4.




        At the low level, there was a significant difference between the




 experimental and theoretical mean values.  At the high level, no




 significant difference between the experimental and theoretical mean




 values was  indicated.




        Based upon the assumption that the theoretical mean values




 represent the true  stack gas velocities,  velocity determinations by




 Method 2 as written provide a reasonable estimate of stack gas velocity




at high velocity levels.  However, at low velocity levels, the method does




not provide a good  estimate of stack gas velocity.




       Since the statistical treatment previously described indicated that




the accuracy and repeatability of Qg appear  to be a direct function of




the same parameters as Vg, the above assessment of accuracy remains




valid for Qs at high and low levels.

-------
                                                                 23
2.     Analysis of Single Traverse Data




       As  shown in the previous section, the limits of accuracy with




which Ap  can be measured at the  individual traverse points will be the




predominant factor in determining  the accuracy of the values obtained




for V ,  over the  range  studied.
     8


       It is believed that since the experimental stack velocity may be




influenced  by both time dependent variations due to changes in process




parameters and by unsymmetrical  flow geometry in the  stack,  a more




accurate estimate of the repeatability of the procedure for determinating




velocity  could be obtained by statistical analysis of the data determined




from traverses of a single diameter.  The use of two diameter traverses




may give a more accurate estimate of the time average  velocity than a




single diameter traverse, but it also has the effect of smoothing the data




since it represents an averaging process.




       Tables  8 and 9 show the data obtained on single diameter pitot




tube traverses at low and high fuel feed rates.  In Table 10 are summarized




the experimental and  theoretical velocities  and volumetric flow rates for




single traverses  at both the high and low fuel feed rates.




       The results of analyzing the single traverse data are shown in




Table 11.  For the experimental stack velocity,  V    ,  since the earlier

                                                 avg


analysis  indicated that  Ap    and  (V Ap)    are the dominant variables
                          avg            avg



in determining precision of the estimate, only these two parameters were




analyzed in the single traverse data.

-------
              TABLE 8 .  VELOCITY TRAVERSE DATA ,
HOUSTON LIGHTING & POWER CO. ,  T.  H.  WHARTON POWER PLANT,
               UNIT NO.  1. LOW FUEL FEED RATE
                    Single Diameter Traverse
    A
                  avg
J. i cl-VCi
Date Run No
5/22/73 1-AE
2 -AW
3-BE
4-BW
5/23/73 9-AE
" 10-AW
" 11-BE
" 12-BW
13-AW
14-AE
DC *-*
. Time 
1125-1135
1145-1155
1210-1220
1225-1235
1025-1030
1035-1047
1055-1108
1115-1130
1135-1145
1151-1200
0.22
0. 33
0.25
0.32
0.23
0.26
0.27
0.26
0.27
0.26
(in ^Ojfc
0.475
0. 573
0.494
0. 547
0.475
0. 521
0.515
0. 510
0.522
0.512
°F
265
266
254
250
252
258
251
250
263
264
M
°R
725
726
714
710
712
718
711
710
723
724
a
in Hg
29. 84
29. 82
29.82
29. 82
29.78
29.77
29.77
29.77
29.77
29.77
%H20
13.9
14. 1
14.4
14.7
13.3
13.6
14.2
13.8
14.5
13.5
%C02
8. 1
8.2
8.4
8.6
7.7
7.9
8.3
8.0
8.6
7.8
%02
6.6
6. 1
5. 1
4. 8
7.0
6.0
5.9
6.6
6.8
5.9
Md
29. 5
29. 5
29. 5
29. 5
29. 5
29. 5
29. 5
29.5
29. 6
29.4
Ms
27.9
27.9
27.8
27.8
28.0
27.9
27.9
27.9
27.9
27.9
Vs
f a.vg scf/hrvirr6
29. 1
35. 2
30. 1
33.2
28. 8
31. 8
31.4
31.0
32. 1
31.4
4. 59
5.52
4. 80
5.29
4. 65
5.07
5.03
4.98
5.03
4.98
                                                                                        CM

-------
              TABLE 9.  VELOCITY TRAVERSE DATA,
HOUSTON LIGHTING k POWER CO. ,  T.  H. WHARTON POWER PLANT,
               UNIT NO. i, HIGH FUEL FEED RATE
                   Single Diameter Traverse
Traverse avS
Date No. Time (inH2O)
5/22/73 5-BW
11 6-BE
11 7-AW
11 8-AE
11 15-AE
11 16-AW
11 17-BE
• " 18-BW
1400-1412
1415-1425
1428-1440
1445-1455
1340-1348
1353-1400
1405-1415
1418-1425
0. 91
0.85
0.84
0.80
0.90
0.88
0.88
0.89
(^>'av
(inH20)^
0.951
0.873
0.911
0.845
0.943
0.930
0.935
0.938
g Ts Ts
°TT °R
a t\.
306 766
308 768
317 777
316 776
286 746
314 774
310 770
310 770
Ps
in Hg
29.80
29.80
29.80
29.80
29.73
29.73
29.73
29.73
%H20
14.9
14.9
14.8
13.8
14.7
14.7
19.6
14.5
%C02
8.8
8.8
8.7
8. 1
8.6
8.6
9.3
8.5
%02
5.2
5.7
5.4
6.2
4.8
5.0
4.6
6.1
Md
29.6
29.6
29.6
29.5
29.5
29.5
29.6
29.6
Ms
27.9
27.9
27.9
27.9
27.8
27.8
27.3
27.9
Vs Q
avg s
fps scf/hr x ID'6
60.0
55. 1
58. 1
53. 6'
58. 9
59.4
60.0
59.5
8.84
8.09
7.54
7.90
8.92
8. 66
8.30
8. 75
                                                                                        CM

-------
                  TABLE 10.  EXPERIMENTAL AND THEORETICAL GAS VELOCITY
                                  AND VOLUMETRIC FLOW RATE -
                           HOUSTON LIGHTING AND POWER COMPANY,
                               T.  H. WHARTON  PLANT, UNIT NO. 1
                                  Single Diameter Traverse
Traverse
No.1
1-AE
2-AW
3-BE
4-BW
5-BW
6-BE
7-AW
8-AE
9-AE
10-AW
li-BE
12-BW
13 -AW
14-AE
15-AE
16-AW
17-BE
18-BW
Fuel Gas2
Feed Rate
MCFH
390
390
390
390
680
680
680
680
385
385
385
385
385
385
680
680
680
680
Load,
megawatts
41. 5
41. 5
41.5
41. 5
71.0
71.0
71.0
71.0
39.0
42.0
40.0
40.0
40.0
40.0
71.0
71.0 "
71.0
71.0
Stack Gas Velocity, fps
Experimental
29. 1
35.2
30. 1
33.2
60. 0
55. 1
58. 1
53. 6
28. 8
31. 8
31.4
31.0
32. 1
31.4
58. 9
59.4
60.0
59.5
Theoretical^
33.3
32.9
31.7
31.0
57. 1
57.2
58. 5
62.7
33.7
33.2
31. 5
32.5
31.4
33.9
56.7
58.8
55.3
59.1
Volumetric Flow Rate, scf/hr x 10"^
Experimental
4. 59
5. 52
4. 80
5.29
8. 84
8.09
7. 54
7. 90
4. 65
5.07
5.03
4.98
5.03
4. 98
8. 92
8. 66
8.30
8.75
Theoretical'1
5.26
5. 19
5. 07
4. 96
8. 44
8. 44
8. 54
9.27
5.46
5. 32
5. 07
5.26
4. 95
5. 39
8. 64
8. 64
7. 99
8. 74
 Letters on run numbers are sample port designation per Figure 1.
^ Fuel gas volume is in cubic feet at 60°F and 1 atmosphere pressure.
^Calculated by the procedure shown in Appendix 1, wet gas basis.
4Calculated by the procedure shown in Appendix 1, dry gas basis.

-------
                           TABLE 11.  STATISTICAL, ANALYSIS, EXPERIMENTAL
                                  AND THEORETICAL STACK VELOCITIES
                                      AND RELATED PARAMETERS,
                                      SINGLE DIAMETER TRAVERSE
EXPERIMENTAL
Percent Distribution About the Mean
Fuel Feed
Rate
Low


High


THEORETICAL
Low


High


Variable Mean
X X
Vs 31.4
^avg °'27
(•/A~)av o.5i
V 58. 1
s
Ap 0.87
favg
(VA~) 0.92
P avg

V 32.5
o/,, c*r*\ Q •>
/o *-'*-' ^ o. £
Ts, °R 717
V 58.2
%C02 8.7
TH, °R 768
Variance
s2
3.58
0.0012
0.0009
5. 82
0.0013
.0014

1.09
0. 102
44. 3
4.91
0. 110
97.3
Standard
Deviation
s
1.89
0.035
0.030
2.4
0.036
0.037

1.05
0. 32
6. 60
2.22
0.34
9.90
2 s x 100
X
95% Confidence Level
12.2
25.8
11. 6
8.3
8.2
8.0

6.4
7.8
1.8
7.6
7.8
2.6
                                                                                                                ISJ

-------
                                                                  28
        As can be seen by comparison of Tables 6 and 11, the results



 are considerably different when single diameter traverse data are



 analyzed and compared to the results obtained from analysis of Method 2



 data.  In Table 11, it can be seen that there is  equality of variance for



 AP    .  When the data from two traverses are combined as  specified
    avg



 in Method 2, equality of variance is not observed due to the smoothing of



 the data. For  the Method 2 analysis, distribution about the mean for



 Ap    is +. 13.5 percent at the low level, while in the single traverse
      6



 data the distribution about the mean for Ap    at the low level is + 25.8
                                         avg                   -



 percent.  However,  at the high  level, the corresponding distributions



 about the mean are + 8.2 percent for single traverse data and +7.4 percent



 for the Method 2 data, indicating that the repeatability of the determination



of  Ap     is better at higher velocities, and little smoothing of data
       avg



occurs by averaging data from two traverses.



        Comparison of the analysis of (VZ^p)   '  indicates that a similar
                                              o


 apparent improvement in repeatability of (yAp)     is obtained by
                                               avg


 averaging the  results of two diameter traverses.  For the Method 2



 results, distribution about the mean for ( T/Ap)    is + 3. 9 and + 43
                                              avg             -  '


 percent at the  low and high levels, respectively, while for the single



 traverse data, the similar values are f 11. 6 and + 8.0, respectively.



 As can be seen,  equality of variance is  not obtained for (-/An)
                                                         ^ avg


 and the variance of (/Ap)    is a function of the level of (//V)
                          avg                                  avg

-------
                                                                 29
       Analysis of Vg    for the  single traverse data indicates that




the distribution about the mean at the low and high levels is + 12.2 and




4- 8. 3 percent, respectively, which corresponds closely with the




distribution about the mean for ( 7A _.)    'at the low and high levels.
                                    P avg



This would be expected from the linear  relationship between Vg    and




( vZAp)a   indicated  by the correlation analysis described earlier.
     tr   o






       Also shown in Table 11 are the results of analysis of the




theoretical velocity,  percent CO^i and T  based on  single diameter




traverse data. The relationships previously described indicating a




major dependence in  the distribution about the mean for V on the




distribution about the mean for percent CO^ still appear valid, with




the apparent repeatability being somewhat poorer as shown by the




increased distribution about the means for the single traverse data




shown in Table 11 when compared to the similar data for Method 2




shown in Table 7.

-------
                                                                  30
3.      Conclusions
        Based upon this study, the following conclusions have been made.




The experimental stack velocity and volumetric flow rate are primarily




dependent upon the accuracy with which  Ap is determined during the




velocity traverse,  over the velocity range  studied.  Under field conditions,




using a Type S pitot tube and inclined manometer, measurements of




 Ap   can be made which provide values of stack velocity with a percent




distribution about the mean of +_ 8. 3 to jf 12. 2,  over the velocity range




studied, based upon single diameter traverse data.  The percent




distribution about the mean for the velocity calculated according to Method 2




is + 4.8 to + 6 . 9 over the velocity range studied.




        Even though velocities calculated according to Method 2 using data




from two diameter traverses 90° apart have a smoothing effect on the data




and shorten  the distribution interval about the mean, when compared to




single  traverse data, accuracy analysis indicates that Method 2 as written




provides reliable estimates of stack gas velocity at  high flow rates, but




that the estimates at low  velocities are unreliable.  This unreliability is




directly attributable to the large variability in  the determination of  Ap




at low  velocities .  Thus,  if greater accuracy or repeatability in determining




the experimental stack velocity is desired, the most profitable area for




improvement would appear to be in improving the accuracy of measuring




 Ap   in the  stack.

-------
                                                                 31
                  METHOD 6 - SULFUR DIOXIDE







      The sulfur dioxide section called for an investigation of possible




variation in collection efficiency with changing concentration over the




applicable range of the method and also for an investigation of the




possibility of low recovery of sulfur dioxide due to retention in the




isopropyl alcohol bubbler.  Accordingly, a special experiment was




designed to  evaluate these effects along with  some other relatively




important factors. The details of the experiment and the results are




given below.





        Collaborative testing of this  method is anticipated at the  facilities




of Walden Research and Monsanto Research, in accordance with plans




and  subcontract arrangements already submitted to the Project  Officer.




At the Dayton power plant, the only control of sulfur dioxide level which




can be obtained is through the addition of dilution air to the  flue  gases.




At Walden,  control of the  sulfur content of the fuel will make it  possible




to obtain varying SO2  levels in the flue gas without dilution.  For this




reason,  a collaborative test at each location is considered preferable.




       Anticipated collaborators include Monsanto Research,  Walden




Research, Southwest Research Institute (Houston laboratory) and




Southwest Research Institute (San Antonio laboratory).  With some




crowding,  a fifth collaborator could be accommodated, and several air




pollution control agencies are being contacted to see if a voluntary




participant  can be obtained.

-------
                                                                 32
        To study the effect of the independent variables on the dependent




variable (observed concentration expressed as a percent of the gas




standard concentration),  a quarter replicate factorial test plan was




designed.  Statistical analysis of the test data should identify which




independent variables  cause significant effects.





        The independent variables studied were the sulfur dioxide




concentration (397 ppm or 707 ppm), the number of sets of  peroxide




impingers in series (one  or  two), the number of isopropyl alcohol




bubblers in series (one or two), the sampling time (20 or 30 minutes),




and the sample volume (0.75 or 1.0 cubic feet).   Variations in sample




flow rate were thus accomplished.  The dependent variable was defined




as the observed concentration  expressed as a percent of the expected




value according to the gaseous standards which were used.   Since the




specific design used investigated seven variables in eight experiments,




and only five were specified, the remaining two were dummy variables.




The dummy variables are unassigned factors and are used to obtain




an estimate of the variance.  The combinations for each experiment




are shown  in Table 12 along  with the value  of the  dependent  variable.




       Before discussing the results, it is important to clarify




the independent variables and to describe the manner in which




intermediate data  were generated.




       The two levels  for the concentration were provided by two




separate cylinders of sulfur  dioxide in nitrogen (397 ppm and 707 ppm)

-------
                                                                 33
      TABLE 12.   FRACTIONAL-FACTORIAL EXPERIMENT DESIGN
        FOR FIVE VARIABLES IN EIGHT EXPERIMENTS SHOWING
                THE VALUES FOR EACH INDEPENDENT
             VARIABLE AND THE DEPENDENT VARIABLE
Experiment    IPA      Peroxide                                  Observed
  Number   Bubblers  Impingers   Concentration  Time   Volume   Recovery
1 2
2 1
3 1
4 1
5 1
6 2
7 2
8 2
2 sets
2
1
2
1
1
1
2
707 ppm
397
397
707
707
707
397
397
20
30
20
20
30
30
20
30
                                                         1.00     99.7
                                                         0. 75
                                                         1.00
                                                         0.75
100. 3
 94. 5
 94. 5
                                                         1.00      90.7

                                                         1.00      97.0

                                                         0.75     102.8

-------
                                                                 34
which were analyzed (using the West-Gaeke method) by the supplier




with accuracy of 0. 5 percent quoted for each.  The  experiments were




conducted shortly after receipt of the standard gases.  Compatibility




with system dynamics was achieved by charging a Tedlar bag from the




respective cylinder and then sampling immediately  from the bag.  A





 commercially produced stack sampling apparatus was not available  during




 the work; therefore, the train was assembled from individual components




 meeting the specifications shown in Figure 6-1 of Method 6 in the




 Federal Register.   The probe and the pitot tube were not required for




 sampling from Tedlar bags.




        For experiment numbers 3 and 5, the bubbler-impinger portion




of the train was identical to Figure 6-1 of the method. In this  config-




uration, measurements  show the vacuum at the suction end of the train




to be 17 to 20 inches of water for flow rates of  1 to  1. 5 liters  per minute.




When two isopropyl alcohol bubblers were called for (experiment




numbers 1, 6,  7,  and 8), an additional bubbler containing 15 ml of




80 percent isopropyl alcohol was inserted in series into the train




following the first isopropyl alcohol bubbler.  When two sets of peroxide-




filled midget impingers were called for (experiment numbers  1, 2, 4, and 8),




two impingers charged with 15 ml of three percent hydrogen peroxide




followed by one empty impinger were inserted in series into the train




following the empty midget impinger.  For example, experiment numbers




1 and 8  contained two midget bubblers charged with isopropyl alcohol

-------
                                                                35
followed by two midget impingers charged with hydrogen peroxide




followed by one empty midget impinger followed by two more  peroxide




impingers  followed by another empty impinger.




       The midget impingers charged with peroxide were treated in sets




containing  two filled impingers and one empty so that the procedure as





described in Sections 4.2 and 4. 3 of the method could be applied to




each set independently.   A result was thus generated and recorded




for each set and the two were added and converted to the percentage




of the expected value to produce the final  result (the dependent variable).




There is a distinct advantage  in this approach since another independent




estimation of collection efficiency can be  made using experiment




numbers 1, 2, 4,  and 8 by comparing the contents of the  second set




of impingers with the first.  These data are  shown in Table 13, and




the results will be discussed subsequently.






       The contents of the isopropyl alcohol bubblers were not




discarded as per the method but were analyzed for sulfur dioxide by




an improvised procedure to oxidize any retained SC>2 to 803 and then




determined by titration as in Section 4. 3 of the method.  The contents




of each  of the bubblers were analyzed separately, and results were




expressed  in terms of percent recovery of the gaseous sample so that




a material balance of the entire train was easily accomplished by




simple addition.   These  percentages were not added into the  dependent

-------
                                                                36
       TABLE 13.   INTERMEDIATE EXPERIMENTAL DATA
                 FROM FRACTIONAL-FACTORIAL
                      EXPERIMENT DESIGN
                                         Peroxide  Impingers
                                           1st      2nd      Total
Experiment
Number
1
2
3
4
5
6
7
8
IPA
1st
1. 1
0. 5
0
0
2.0
1.0
0. 5
0.8
Bubblers
2nd
1.7
-
-
-
-
0. 1
0. 5
2. 5
                                          91.7

                                          99.0

                                         100.3

                                          94.5

                                          94. 5

                                          90.7

                                          97.0

                                         102.8
0.2
0.7
 0
 91.9

 99.7

100.3

 94. 5

 94. 5

 90.7

 97.0

102. 8
#Numbers represent percentage of expected value based on concentration
  of gas standard.
variable.  The data are shown in Table 13.  In order to confirm that

the contents of the bubblers represented retained SO2 rather than any

SO, present in the sample, a  series of experiments was run in which

the bubbler contents were not oxidized but rather titrated directly so

that any sulfur oxides detected would be attributable to SO3 rather than
retained SO£.  The results (not shown) showed no detectable SO3 in

any of the bubblers in any of eight runs.  Therefore,  the respective

results  shown in Table 13 are assured to be due to SO2 retention.

These results will be discussed in more detail later.

-------
                                                                37
       The two levels of time and sample volume are straightforward.




The actual selection of the times and volumes are based on the minimum




time and minimum volume as designated in the Federal Register under




Section 60.46(c)(2), Test Methods and Procedures.   The various




combinations of sample time and sample volume produce flow rates




of 0.71,  0.94,  1.06,  and 1.42 liters per minute.





       Mechanical difficulties were encountered throughout the experi-




ments,necessitating the repeating of some of the runs one or more times.




These difficulties were associated with  pump leakage, meter malfunction,




or loose connections.   It is, therefore,  very important to make the leak




check as described in Section 4. 1. 1 of the method.




       Another difficulty encountered was in the carryover of isopropyl




alcohol which often occurred at sample  flow rates which were




about 1.4 liters per  minute   (1 cu ft in 20 min) such as in experiment




numbers 4 and 7.




       The results of the statistical analysis are given in Table 14.




The table shows the net effect  [the difference between the average value




of responses at the high (+) level and the average value of responses




at the low (-) level] for each variable.  The significance of each effect




(with  respect to the dependent variable)  is indicated by the absolute value




of the t-statistic and the corresponding  percentage  probability.  The




significance percent for each effect thus provides an estimate of the




probability of finding an effect that large due to chance or experimental

-------
                                                               38
             TABLE 14.  STATISTICAL, ANALYSIS OF
                FRACTIONAL-FACTORIAL DESIGN
                        jC
 Youden Ruggedness Test


 Plackett-Burman Design for 7 Factors and 8 Experiments'


 System - Method 6


 Response - Recovery
Variable
No.
1
2
3
4
5
6
7
Variable
Name
Dummy
IPA Bubblers
Impinge rs
Dummy
Concentration
Time
Volume
Effect
(- to +)
-1.30
-1.65
-1.60
1.55
-7.05
1.00
-1.90
|t| -Value
0.85
1.08
1.05
1.02
4.62
0.65
1.24
Significance
Percent
42
32
34
35
0.8
>50
25
Average value  of response = 96.425


Standard error = 1. 5266


Degrees of freedom = 2
* Youden, W.  J. ,  "The Collaborative Test,"!, of the A.O. A. C. ,  46,  No.  1,
(1963), pp 55-62.
 Plackett, R. L. and Burman, J.  P. , "Design of Optimum Multifactoral
Experiments," Biometrika,  33, (1946),  pp 305-325.

-------
                                                                 39
error alone.  If the effects of one or more dummy variables were




significant, there would be either (1)  significantly large interactions




of main effects,  (2) important independent variables omitted or not




held constant,  or (3) considerable error in the measurement technique.




       The only  effect sufficiently large to be significant (10 percent




level of significance) is the concentration.  The effect is negative,





indicating a decreased response in going from the low level to the  high




level.  In the absence of data from the second set of peroxide impingers




from experiment numbers 1 and 4,  this might be interpreted as a




decrease in collection efficiency at the higher concentration level.




Since the second set of impingers contains little or no sulfur dioxide




(see also experiment numbers 2 and 8 at the lower concentration level),




plus the fact that there is not a sufficiently significant effect from




peroxide impingers (the effect is  even in the wrong direction), the




natural conclusion  is that the higher concentration is very probably




in the neighborhood of 660 ppm.  The importance of retrieval of inter-




mediate data now becomes evident.




       To  summarize the results thus far:  there are no significant




effects due to isopropyl alcohol bubblers,  peroxide impingers,  time,




volume, or dummy variables.  The effects of concentration are either a




concentration bias in the method or an inaccuracy in the concentration




of the 707 ppm cylinder.  Unfortunately, at the time of report preparation,




the 707 ppm standard was no longer available for independent analysis.

-------
                                                                  40
        The effects of sulfur dioxide retention in the isopropyl alcohol




 bubblers are negligible from the analysis above and are also minor





 according to the data  in Table 13.   The average retention in a bubbler




 is less than  one percent of the amount present (average of 0. 9 percent




 for 12 observations).   The values  can be seen to range from zero to




 2. 5 percent at the highest. All  of these results were  obtained using




 the 15-minute purge as specified in Section 4. 1.2 of the method.  A




 single experiment (under the  same conditions as experiment number  1)




 in which the  purge was not done  showed 7 percent retention in each of





 the two bubblers for a total of 14 percent.   It is, therefore,  quite




 important to follow the purging procedure rigorously.





        The minimum  detectable limit (based on a net  titration of 0. 1  ml,




a 0. 75 ft   sample, and a 10 ml aliquot) is 3 ppm which should cause




no limitations  in the use of the method.  If more sensitivity  was desired,





a larger sample could be taken.  The method can conveniently analyze




samples up to  1400 ppm (based on 50  ml titration, 0.75 ft  sample,




10 ml aliquot),  providing the collection efficiency does not deteriorate




at that level.

-------
                                                               41
                METHOD 7 - NITROGEN OXIDES







        Laboratory work was conducted to check out the entire procedure




 and especially to investigate the reported interference from hydrogen




 chloride.  A total of 64 experimental tests were made using standard




 nitric oxide mixtures of 98 and 700 ppm.




        In 48 tests comprising five sets of experiments, the 98 ppm




 standard gas was used as a test gas in flasks which were spiked with




 hydrogen chloride of known concentration.  Concentrations of hydrogen




 chloride were  11 ppm,  50 ppm,  100 ppm, 500 ppm, and 1120 ppm.




 In two sets, the spiking was done using hydrochloric acid of known




 concentration while in the other three sets,  spiking was accomplished




 by  injecting the proper amount  of dry hydrogen chloride gas.





       In 16 other tests,  the 700 ppm test gas was used, half without




hydrogen chloride and half with the addition of sufficient hydrochloric




acid to give a hydrogen chloride concentration of 700 ppm.  All




samples were analyzed according to the  procedures described in




EPA Method 7.





       Table 15 and Figure 2 presents the  results obtained.




Based on these results and the complete laboratory data, the following




conclusions are established concerning Method No. 7:




       1.  The method is tedious and time consuming, especially in




       the analytical phase.  This, of course, was known previously




       and has been the subject of some discussion and comment.

-------
                                                           42
TABLE 15.   DATA TABULATION—INTERFERENCE OF HYDROGEN

                                                          7
CHLORIDE WITH NO  DETERMINATION - EPA METHOD NO.
                  X.
                                                 ANALYSIS
HC1
Cone, by Sample
Chloride Vol. Serial
Date Set Source PPm No.
11/16/72 1 hydrochloric 1120 1
acid 2
3
4
5
6
7
8
Avg.
11/20/72 2 hydrochloric 11 9
acid 10
11
12
13
14
15
16
Avg.
11/21/72 3 dry hydrogen 500 17
chloride 18
19
20
21
22
23
24
11/27/72 500 25
26
27
28
29
30
31
32
Avg.
Stand.
NO
ppm
Vol.
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
Spiked
NO
Cone.
ppm

--
__
--
15
20
20
__
18
__
--
--
-_
95
97
94
97
96
61
65
52
56
73
83
__
61
_ _
_ _
--
--
69
62
55
62
64
Unspiked
NO
Cone.
ppm
110
121
-.
116
__
__
--
—

103
112
108
105
--
--
--
--

w M

_ _
_ _
__
_ _
_ _
_ ..
102
100
97
95
--
- -
..
_ _


-------
                                                             43
Table 15.   Data Tabulation--Interference of Hydrogen Chloride
   with NO  Determination - EPA Method No. 7 (Cont'd.)

                                                 ANALYSIS
HC1
Cone, by Sample
Chloride Vol. Serial
nate Set Source ppm No.
jl/29/72 4 dry hydrogen 100 33
chloride 34
35
36
37
38
39
40
Avg.
12/5/72 5 dry hydrogen 50 41
chloride 42
43
44
45
46
47
48
Avg.
7^/6/72 6 hydrogen chloride 700 49
50
51
52
53
54
55
56
iZ/H/72 hydrogen chloride 700 57
1 58
59
60
61
62
63
64
Avg.
Stand.
NO
ppm
Vol.
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
98
700
700
700
700
700
700
700
700
700
700
700
700
700
700
700
700
700
Spiked
NO
Cone.
PPm
--
- -
--
__
83
82
79
82
82
--
--
—
91
88
89
89
89
--
--
--
447
412
502
469
--
--
--
--
--
429
521
409
456
Unspiked
NO
Cone.
ppm
89
f\ 4
91
91
90
--
--
--
--

91
92
92
90
--
--
--
--

943
841
917
847
--
--
--
--
602
--
712
545
--
--
--
«• ^


-------
                                           Actual NOX concentration in
                                                all runs = 98 ppm

                                                     o Cone,  as NO  ppm
 200
400
 600         800
HC 1   ppm
1000
                                                           1200
1400
Figure 2.  Interference of HC1 With the Determination of NOX —
              EPA Method No. 7, Average Points

-------
                                                                 45
       2.   Results are affected by the presence of hydrogen chloride,




       as has been indicated by previous publications.  The degree of





       interference is approximately linear with hydrogen chloride




       concentration,  ranging as high as  78 percent with





        1120 pprn    HC1 when sampling a test gas of approximately




       100  ppm NO .   A similar relationship appears to apply in the




       case of the 700 ppm test gas,  although tests were only run at a





       single level of HC1, 700 ppm.  At this level, the indicated NOX




       level is about 65 percent of the original value as determined




       in tests without hydrogen chloride.   It should be noted that





       erratic results were obtained  with the unspiked 700 ppm samples.





       The  sensitivity of the  method along with  the minimum detectable




limits for NOX by this  method have been estimated.  Data used in





the calculations of the minimum detectable limit were taken from a




calibration  curve constructed by analysis  of a series of standard





solutions of potassium nitrate.   Concentration range of the standards was




zero to 400  (j.g nitrogen dioxide.  Measured absorbance of the solutions




was in the range of 0-0. 5  absorbance units, using absorbance cells of




1.2 centimeter path length.   The calculations are presented in Appendix III.





The minimum detectable limit is estimated to be 2 ppm NOX as NCK in




the gas sample.  The upper limit without dilution is about 100 ppm




NOX as NO2 in the gas sample.  To analyze gas samples  containing




around 700  ppm NOX would require a  tenfold dilution according to Section 4. 3. 1




of the method (Appendix III,  Figure 1).

-------
                                                               46
       The maximum sensitivity of the method can approach 0. 2 ppm




NOX provided the absorbance can be read to 0.001 absorbance units,




but in a practical sense would probably be between 0. 2 and 2. 0 ppm NOX.

-------
                                                      I-i


                        APPENDIX I


         Glossary of Terms and Equations



Terms and equations taken from The Federal Register,


Volume .36, December 23,  1971:


Equations
(V8)a    =  Average Stack Gas Velocity, ft/sec
Q       =  Stack Gas Volumetric Flow Rate,  ft /hr
Qs       = 3600 (1 - B   >  ,=-    Jbr     VSA
Terms
       = Conversion constant : K  =  85.48
                                           sec   L/b-mole



       = Pitot tube coefficient, dimensionless
P


       = Average velocity head of stack gas
  avg
                                                             ,

                                                           g— )
                                                            K/
            L        it  i    /\  ,  where   >1 = number of traverse points
            in       'l /	*—*P;
             ayg      i~T7

 /^\  \  = Average of the square roots of the velocity head
 \   P/avor     measurements taken at individual traverse points

                        i =n

         "~)
               /avg
           (J/\  1    = _L \ v^p.»  where lr\ = number of traverse points
           \   P/avg   n  '      i
                          i = 1

-------
                                                                1-2
        (Ts)avg =  Average absolute stack gas temperature,  R
                              I
                   (Ts)    =  —- y *• i, where  y\ = number of traverse points
                         5       f   i ••
                                 i = 1
        Tstd    = Absolute temperature at standard conditions, 530°R
                =  Absolute stack gas pressure, inches mercury
        P . ,     = Absolute pressure at standard conditions, 29.92 inches
                      mercury
        B       =  Proportion by volume of water vapor in the stack gas
                = Dry molecular weight of the stack gas (calculated from
                      Orsat analytical data)
       M       = Molecular weight of the stack gas, wet basis


                  M   =  M, (1 - B   ) +  18 B
                    s      d x     wo'       wo


B.     Terms and equations used in statistical analysis of data:
                               	       'i - y)
                               i  =  i    	
Correlation Coefficient  =
                                      y (
                                 ,/
                                    i = i
                                                  _ 2
                                              ^ - y)
               where x is the independent variable and y is  the
                   dependent variable,  i.e.  y = f(x)

-------
                                                              1-3
       x is the arithmetic mean of variable x
           x   = —  \  xj,  where  )^ is the sample size



                     i  =  1


        2                                                    xt-'
       s  is the unbiased  estimate of the population variance,  O
                              ' -^ - x)  ,  where x and )\ as above
                           i = 1
        is the  sample standard deviation
            8  =
       Percent distribution about the mean, 95 percent confidence level


          is defined as





                        2s •  100       , .    ,
                        	 , s and x as above
                           x
       t-value is a  student's t with six degrees of freedom





                        t =      effect

                           standard error




C.     Conversion  Factors:




           inches x 2. 540 = centimeters




           cubic feet/unit time x 0.0283 = cubic meters/unit time




           feet per second x 0. 3048 = meters per second




           pounds/unit time x 0. 4536 = kilograms/unit time




           feet x 0. 3048 = meters

-------
                                                               II-l




                              APPENDIX II





          DATA FROM T. H. WHARTON POWER PLANT,

           HOUSTON LIGHTING AND POWER COMPANY






Sample calculation of theoretical stack velocity and volumetric flow rate.




Carbon in fuel gas,  calculated from gas chromatographic analysis:  1.072 Ib-



                             ^f
atom carbon/lb-mole fuel gas.  . Carbon in stack gas,  from Orsat Analysis




Run 1-AE, Table  1:  0.081 Ib-atom carbon/lb-mole stack gas.




        mole dry stack  gas/mole fuel gas in:





            i-072  = 13.24

            0.081



        moles H2O  = 2 x moles CO2  = 2x0.081  =  0.162




        Total moles stack gas/mole fuel gas in:




            13.24 + (0.  162)(13. 24)  =  1.162x13.24  =  15.38




        Theoretical  gas velocity, wet basis:



                                                        3              T
            Total  moles stack gas       Fuel feed rate,  ft /hr      _      s
                                 X
              moles fuel gas in     (Stack area,  ft^)(3600 sec/hr)    T,   ,
                                                                      XX16J. £cl S





                         moles stack gas         390 x JO3 ft3/hr        725°R

                   5'38  moles fuel gas in X  (69.81 ft2)(3600 sec/hr)  X 520°R





                =  33.3 fps




       Theoretical volumetric flow rate, dry basis:




            moles dry stack gas    c  , ,  ,    .   ,^3 ,.        530°R
            	•	<-——:—a— x  fuel feed  rate, ft^/hr x 	~
               moles fuel in                               Tfuel gas




            where:




            530°R  =  70°F  =  EPA specified  temperature



            13>24  moles dry stack gas  x  39Q x  1Q3 ffc3/hr x  530^

                   moles fuel in                             520 R





                =  5.26x 106ft3/hr.
* 96.2% methane.

-------
                                                               ni-i



                           APPENDIX III






             Estimation of Limits for NOX by Method 7






       The approximate minimum detectable limit for NO  by Method 7




can be estimated from laboratory data and the appropriate equations for




calculating the concentration of nitrate in the samples.




       Ref. :  Method 7, Section 6.2,  Federal Register, Vol. 36,




       p. 24893, 23 December  1971.





                           lb

       r  _   m         cuft      \  _/.,..   -5 Ib/scf U m

       c  -
       where C  =  cone, of NO  (as NO-,), Ib/scf
                              x        £
              m  =  mass of NO£ in gas sample,  |o.g




            V   =  sample volume at standard conditions, ml.
              sc       r



       V    =  200 ml
         sc



       According to Sections 3.3.4 and 5.2,  the calibration curve is




prepared using diluted standard solutions with a concentration range of




0 ^ m ^ 400  (jig.  The resulting calibration curve in our laboratory gives




806 [ig NO^/absorbance unit,  using absorption cells of 1.2 cm path length.




Therefore,  an absorbance of 0. 010 is equivalent to 8. 06 (xg NO2 and
C  =   -2000-      = 25 x 10-8  = 0.25 x 10-6






          lb     359 ft3/r^mole      530°R       10  ft
                                               °           6  3


                         46 lb/l>.-.nuie  X   ^92°R     million ft3




       ppm  = C x (8.4 x 106)




       ppm  = (0.25 x 10-6) x (8.4 x 106)  =  2. 1 ppm




       A concentration of 2  ppm NO£ in the gas sample is probably




detectable.

-------
                                                              ni-2
       A maximum sensitivity of about 0.2 ppm is possible if absorbance




can be measured to 0. 001 absorbance unit.




       As shown in Figure 1, the upper limit of gas sample NOX




concentration without dilution is about 100 ppm based on an absorbance of




0. 5 unit.

-------
                                                               Ill-3
   1.0
u
c

-------
                                                               IV-1


                            APPENDIX IV



                       Accuracy of Method 2




       To assess the acuracy of the experimentally determined



Method 2  stack gas velocities with respect to the theoretical stack gas


velocities, a Student's t-test is used to obtain a significance level.




       Assume Vtheoretical represents the true stack gas velocity.




Then,  using a calculated t-value



             V"         V
       tc =   savg "   theoretical
                  s
       where s is the sample standard deviation of Vs    and >^
                                                       O

       is the number of observations of Vc    used to obtain Vs
                                         savg                  avg



       The significance level resulting from the test will be  the probability


of obtaining a sample (experimental) mean Vs    which differs from the
                                            avg

true (theoretical) mean %eoreUcal by a magnitude of | V^  - Vtheoretic


or greater due to chance alone.  The probability is obtained from a table


of Student's t distribution with ( ^- 1) degrees  of freedom. A low



significance level,  less than 10 percent, indicates that v*0    is not a
                                                        savg

good estimate of Vtheoretical.




       1.       Low Level


                The hypothesis to be tested is  that JJL = 32. 5 (Vfcn     .   .).



From Table 6 ,  Vg    =31.4 with a sample standard deviation s = 0.756.
                   avg

-------
                                                               IV-2
 Then,

                       31.4 - 32. 5
                 c  7  (0. 756)/


                       -1. 1
                 tc  " 0.337


                 tc  = -3. 264 with 4 degrees of freedom

 From the Student's t table, the percent significance level for tc = -3.264

 with four degrees of freedom is 3. 49%.


        2.        High Level

                 The hypothesis to be tested is that p. = 57. 4 (Vtheoretical).

 From Table 6 ,  Vg    = 58. 1 with a sample standard deviation s =2.015.
                   avg
 Then,
                      58. 1-57.4
                tc =
                      (2.015)/

                       0.7
                      1.0075
                tc  =  0. 695 with 3 degrees of freedom

From the Student's t table, the percent significance level for tc  =  0. 695

with three degrees  of freedom is 56. 08%.

-------
TECHNICAL REPORT DATA
_ (Please read liiitnii-tiiins on //.v /vrt rsc before eomi'lciiiii;)
•• REPORT NO 2
EPA-650/4-74-039
4. TITLE AMD SUBTITLE
"Laboratory and Field Evaluations of EPA Methods
2, 6, and 7."
'. AUTHOR(S)
Henry F. Hanril
*. PERFORMING ORGANIZATION NAME AND ADDRESS
Southwest Research Institute, 8500 Culebra Rd.,
San Antonio, Texas 78284
	 	
|!Z. SPONSORING AGENCY NAME AND ADDRESS
Environmental Protection Agency, NERC, QAEML
Methods Standardization & Performance Evaluation Branch
Research Triangle Park, N. C. 27711
S. SUPPLEMENTARY NOTES
3. RECIPIENT'S ACCESSION-NO.
5. REPORT DATE
October 1973
6. PERFORMING ORGANIZATION CODE j
8. PERFORMING ORGANIZATION REPORT NO. I
SWRI Project 01-3462-001
10. PROGRAM ELEMENT NO.
1HA327
11. CONTRACT/GRANT NO.
68-02-0626
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE

 6. ABSTRACT 	!	~~	~~"	'	~	
           A study was made to evaluate Methods  2,  6,  and  7,  proposed by EPA for
 determination of stack gas. velocity  and volumetric flow rate,  sulfur dioxide emissions,
 and nitrogen oxide emissions.  These evaluations were conducted prior to collaborative
 testing of the subject methods. Method 2: Stack Gas Velocity and Volumetric Flow Rate:
 Statistical analysis  of stack gas velocity data indicates tnat Method 2 provides an
 accurate estimate of the true stack  gas velocity at high  gas velocities. Accuracy of   J
 Method 2 velocity estimates at low gas velocities  is  shown to  be unreliable.  Correlation
 analysis demonstrates that the volumetric flow  rate estimates  have  the same character- j
 Tstics as  the velocity estimates. Correlation analysis  also  demonstrates that the      I
 variation  in the stack gas velocity  and volumetric  flow rate estimates is principally  j
 due to variation in determination of Ap, the velocity head in  the stack.  By way of
 ^mparison, a separate analysis was  performed on individual  velocity traverse data.     |
      ' 6 -  Sulfur Dioxide: Investigation of possible  causes  of variation in collection
            of S02 in Method 6 was made. Concentration of SO?  in  the  stack  gas  is  shown
 o be  the  only factor to have any significant effect on col lection efficiency.  The
Purge  period  specified in Method 6 was shown to be necessary  to  avoid apparent low  S02
values due' to retention in the isopropanol bubbler. The minimum  limit is estimated  to
Je 3 ppm.  Method  7-Nitrogen Oxides; Investigation of possible interference with NO
germination bychloride ion indicated the degree of interference to be-linearly  re*-
                         >eerttration.
                              KEY WORDS AND DOCUMENT ANALYSIS
                DESCRIPTORS
DlsTRlBUTION STATEMENT


  limited
                                            b.lDENTIFIERS/OPEN ENDED TERMS
  COSATI Held/Group
                                            19. SECURITY CLASS (/VIM Keport)
                                             Unclassified	.
21. NO. OF PAGfcS

      62	
     2220-1 (9-73)
                                            20. SECURITY CLASS ( Hiis page)

                                             Inclassified	
                                                                      22.

-------