United States                   EPA-600/7-85-024
Environmental Protection
                         June 1985
Research and
Development
COAL SAMPLING

AND ANALYSIS:

METHODS AND MODELS
Prepared for
Office of Air Quality Planning and Standards
Prepared by
Air and Energy Engineering Research
Laboratory
Research Triangle Park NC 27711

-------
                  RESEARCH REPORTING SERIES


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

    1.  Environmental Health Effects Research

    2.  Environmental Protection Technology

    3.  Ecological Research

    4.  Environmental Monitoring

    5.  Socioeconomic Environmental Studies

    6.  Scientific and Technical Assessment Reports (STAR)

    7.  Interagency  Energy-Environment Research and Development

    8.  "Special" Reports

    9.  Miscellaneous Reports

This report has been assigned to the INTERAGENCY ENERGY-ENVIRONMENT
RESEARCH AND DEVELOPMENT series. Reports in this series result from the
effort  funded under the 17-agency Federal  Energy/Environment Research and
Development Program. These studies relate to EPA's mission to protect the public
health  and welfare from adverse effects of pollutants associated with energy sys-
tems.  The goal of the Program is to assure the rapid development of domestic
energy supplies in an environmentally-compatible manner by providing the nec-
essary environmental data and control technology. Investigations include analy-
ses of  the  transport of energy-related pollutants and their health and ecological
effects; assessments  of, and development of. control technologies for energy
systems; and integrated assessments of a wide range of energy-related environ-
mental issues.
                        EPA REVIEW NOTICE
This report has been reviewed by the participating Federal Agencies, and approved
for  publication. Approval does not signify that the contents necessarily reflect
the views and policies of the Government, nor does mention of trade names or
commercial products constitute endorsement or recommendation for use.

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

-------
                                      EPA-600/7-85-024
                                      June 1985
            COAL SAMPLING AND ANALYSIS:

                 Methods and Models
                        By

     Alan Glelt, William Moran and Arthur Jung
                    VERSAR  INC.
                   P.O.  Box 1549
                 6850 Versar Center
            Springfield, Virginia  22151
    Contract Number:  68-02-3181, Tasks 10 and 13
         James 0. Kllgroe,  Project Officer
        U.S. Environmental  Protection Agency
Air and Energy Engineering  Research Laboratory
   Research Triangle Park,  North Carolina  27711

                   Prepared for:

U. S. ENVIRONMENTAL PROTECTION AGENCY
      Office of Research and Development
             Washington. DC  20460

-------
         COAL SAMPLING AND ANALYSIS:

              Methods and Models
                 Prepared by:

  Alan Glelt, William Moran and Arthur Jung
                 VERSAR INC.
                P.O. BOX 1549
              6850 Versar Center
         Springfield, Virginia  22151
Contrac.$ Number:  68-02-3181, Tasks 10 and 13
                Prepared for:

      James D. Kllgroe, Project Officer
     U.S. Environmental Protection Agency
 Industrial Environmental Research Laboratory
Research Triangle Park, North Carolina  27711
                Date Prepared:

                   May 1984

-------
                                 ABSTRACT

     New Source Performance Standards for large coal-fired boilers and
certain State Implementation Plans require operators to monitor SOp
flue gas emissions.  As an alternative to stack gas monitoring, sampling
and analysis of feed coal has been proposed to estimate S02 emissions.
This report provides Information on coal sampling and analysis (CSA)
techniques and procedures and presents a statistical model for estimating
SO  emissions.  In particular, this study assesses the various coal
sampling techniques and equipment, the various sample preparation and
analytic methods, and common practices for coal sampling and analyses.
It describes the variables associated with the prediction of SO-
emissions from CSA data such as sulfur retention, variability, measuring
errors, and auto-correlation.  Finally, 1t presents a time series model
for predicting emissions which takes Into consideration the correlation
of the sulfur content of the coal, the measuring errors, and the sampling
procedures for coal collection.  The model 1s used to fit 53 data sets
with little evidence of non-fit.

-------
                             ACKNOWLEDGEMENTS
                                                        *
     The support of the U.S. Environmental Protection Agency,  Industrial
Environmental Research Laboratory,  Research Triangle Park,  North Carolina
1s gratefully acknowledged.   We are particularly Indebted  to James D.
Kllgroe for his cooperation, active support and sustained  Interest 1n  the
project.
     We should also like to  thank Thomas Logan and Louis Paley (EPA);
Ross Leadbetter (Univ. North Carolina);  and James Peeler (Entropy
Environmentalists) for their support and guidance to the project.
Special thanks go to B. Woodcock and S.  F1gg1ns (Versar) for their
computer Implementation of the model described 1n Chapter  8.

-------
                            TABLE OF CONTENTS


                                                                 Page No.
1.   Introduction 	      1

2.   Conclusions and Recommendations 	      5

3.   Assessment and Statistical Evaluation of Coal Sampling
     Techniques and Equipment 	      6

     3.1   Objectives of Coal Sampling 	      6
     3.2   Sampling Guidelines 	      8
     3.3   Sampling Equipment 	     14
           3.3.1  Manual Sampling 	     14
           3.3.2  Mechanical Sampling 	     19
     3.4   Sampling Considerations	     28

4.   Assessment and Statistical Evaluation of Sample
     Preparation and Analytical Methods 	     32

     4.1   Sample Preparation 	     32
           4.1.1  A1r Drying 	     33
           4.1.2  Sample Reduction 	     35
           4.1.3  Sample Division 	     35
           4.1.4  Sample Reduction for Analysis 	     45
           4.1.5  Sample Mixing 	     45
     4.2   Sample Analysis 	     45
           4.2.1  Determination of Moisture 	     48
           4.2.2  Ash -J)eterm1 nation 	     52
           4.2.3  Determination of Total Sulfur 	     54
           4.2.4  Determination of Calorific Value 	     60

5.   Sulfur Loss 1n Coal-Fired Utility Boilers 	     63

     5.1   Sulfur Loss In Pulverizers 	     64
     5.2   Sulfur Loss 1n Combustion 	     64
     5.3   Guidelines 	     65

6.   Common Coal Industry and Utility Practices for
     Coal Sampling and Analysis 	     66

     6.1   Information Sources 	     66
     6.2   Sampling Techniques 	     67
           6.2.1  Type I 	     67
           6.2.2  Type II 	     68
           6.2.3  Survey Results 	     68
                                     IV

-------
                            TABLE OF CONTENTS
                               (continued)
                                                                 Page No.
     6.3   Sampling Location 	     70
     6.4   Bias Testing/Sampling Quality Control 	     71
     6.5   Analytical Procedures 	     71
           6.5.1  Sulfur Analysis 	     71
           6.5.2  Heat Content, MosHure, Ash Analysis 	     73
     6.6   Analytical Quality Control 	     75
     6.7   Summa ry 	     75

7.   The importance of Coal Sulfur Parameters 	     77

           7.1  Time Series 	     77
           7.2  Coal Time Series 	     78
           7.3  Variability of Coal  Parameters 	     78
                7.3.1  General Characteristics 	     79
                7.3.2  Importance of Variability 	     79
                7.3.3  Reducing Coal Sulfur Variability 	     81
                7.3.4  Measuring Error Variability	     81
           7.4  Auto-Correlation of  Dally Measurements 	     84
                7.4.1  Correlation 	     84
                7.4.2  Auto-correlation 	     86
                7.4.3  Auto-correlation 1n Coal  	     86
           7.5  Variance and Averaging Period 	     87

8.   The Model 	     91

     8.1   Explanation of a CEM System	     92
     8.2   Explanation of a Method 6B/Cont1nuous Bubbler
           Monitoring System	     92
     8.3   Description of the Model  	     93
     8.4   Estimation of the Measured Dally Composite 	    100
     8.5   Example 	    104

9.   Considerations of Measurement Error 	    107

     9.1   Sampling Error 	    107
     9.2   Sampling Location Considerations 	    108
     9.3   Averaging Time and Lot Size 	    110
     9.4   CSA Quality Assurance/Quality Control 	    Ill
           9.4.1  Sampling 	    Ill
           9.4.2  Sample Preparation and Analysis 	    112

-------
                            TABLE OF CONTENTS
                               (continued)
                                                                 Page No.
 10.  Considerations of Sampler Bias 	     113

     10.1  Bias Test Procedures 	     113
     10.2  Size Bias of Automatic Samplers 	     114
     10.3  Sulfur and Ash Bias Simulation 	     116
     10.4  Theoretical Calculations 	     119
     10.5  Bias Testing Using Emissions Monitors 	     128

 11.  Data Analysis 	     137

     11.1  The Data 	     137
     11.2  A Typical Analysis 	     138
     11.3  Goodness-of-F1t Analysis 	     139
     11.4  Verification of the Coal Model 	     140
     11.5  Comparison of the Models for CSA and CEM 	     145
     11.6  Analysis of the R&F Data 	     147
     11.7  Measurement Error Analysis  	     147
     11.8  Conclusions 	     150

 12.  Correlation of CSA Data to S02 Emissions 	     151

 References 	     152

Table of Conversion Factors	    155

Appendix:  Data Sets.."	    A-l

-------
                              LIST OF FIGURES


                                                                  Page No.
 3-1   Holmes  Sample Probe 	      17
 3-2   Sample  Probe 	      17
 3-3   Prelser Hand Sampler 	      17
 3-4   Ramsey/Gal1gher Primary Sampling Machine 	      20
 3-5   Denver  Equipment Division's  Chute Sampler 	      22
 3-6   Quality Control Equipment Company's  Cross Cutting
      Sampler 	      23
 3-7   McNally Belt Sampler 	.!	      23
 3-8   McNally Plate Sampler  	      24
 3-9   Harrison R.  Cooper  Systems,  Inc. Traversing Cutter 	      25
 3-10 Harrison R.  Cooper  System's  Rotary Scoop Sampler 	      26
 3-11 Bristol Engineering Company's  Rotary Arm Scoop 	      27

 4-1   A1r Drying  Oven	      34
 4-2   Jaw Crusher  	      36
 4-3   Hammermlll  Crusher  	      37
 4-4   Rotary  Crusher 	      37
 4-5   Riffle  	      38
 4-6   Rotary  Plate Divider	      40
 4-7   Ramsey  Sampler/Gal1gher Terltlary Rotary 	      41
 4-8   Denver  Vezln Sampler 	      42
 4-9   McNally Rotary Drum Sampler  	      43
 4-10 Ramsey  Three Stage  Sampling  System 	      44
.M-ll Braun Lab Pulverizer 	      46
 4-12 Sturtevant  Sample Grinder 	      46
 4-13 Small Lab Riffle.	      47
 4-14 Mixing  Wheel 	1	      47
 4-15 Forced  A1r  Oven 	      50
 4-16 Muffle  Furnace 	      53
 4-17 AdlabatU Bomb Calorimeter 	      57
 4-18 Tube Furnace 	      57

 6-1   Frequency of Sampling  Method 	      69
 6-2   Summary of  Sampling Locations  	      72
 6-3   Frequency of Sulfur Analysis Method  	      74

 7-1   Effect  of Variance  on  S02 Emissions  	        80
 7-2   Republic Steel Corporation - Hourly  Increment
      Data for Total Sulfur  	      82
 7-3   Republic Steel Corporation - Hourly  Increment Data for
      Sulfur  Dioxide Emission Rate 	      83
 7-4   Variability  of Data 	      85
 7-5   Coal Data with r=.7.  Exceedances exist	      88
 7-6   Coal Data with r=.l.  No exceedances	      89
 7-7   Effect  of Averaging Period on  Compliance and Variance 	      90
                                    Vll

-------
                             LIST OF FIGURES
                               (continued)
8-1  Increment Collection and Composite 1n a CSA Program
8-2  Increment Collection and Composite 1n a CEM Program
8-3  Increment Collection and Composite 1n a CB Program .
10-1 Size Distribution of Reference Sample ..
10-2 Size Distribution of Test Sample 	
10-3 Size Fraction vs. Sulfur Content 	
10-4 Affect of Bias on Relative Accuracy Test
Page No,

   94
   95
   97

  120
  122
  125
  127
                                   Vlll

-------
                              LIST OF TABLES


                                                                 Page No.
3-1  Summary of Sampling Guidelines 	      15
22  Order of Preference of Sampling Procedures and Methods  ...      29

10-1 Size Distribution of EPRI Bias Test Data 	     115
10-2 Helvetia Coal Mine Data	     117
10-3 Lower Freeport Coal Data 	     118
10-4 Size Fraction vs. Sulfur and Ash Content 	     124
10-5 Biased Sampler Sulfur Content 	     126
10-6 Probabilities of CSA Passing Bias Test Relative to
     Method 6B 	     132
10-7 Probabilities of CSA Passing Bias Test Relative to CEM  ...     133
10-8 Probabilities of CSA Passing Bias Test Relative to
     Method 6B (Alternate Depletion) 	     135
10-9 Probab1lt1es of CSA Passing Bias Test Relative to CEM
     (Alternate Depletion) 	     136

11-1 Summary of Results for CSA 	     141
11-2 Summary of Results for CEM 	     142
11-3 Summary of Results for Method 6B 	     143
11-4 Comparison of CSA and CEM for Homer City (Time 1) 	     146
11-5 Comparison of R&F Split Data Sets 	     148
11-6 ANOVA for Paired R&F Data 	     149

-------
Table of Conversion Factors
Multiply
English Unit
pound (Ib)
ton (2000 Ib)
inch ( in . )
foot (ft)
British Thermal Unit
(Btu)
Btu/lb
lb/106Btu
gal Ion /minute (gpm)
Tyler Screen Size Mesh
Mesh Size
14
24
48
100
200
270
325
2L
453.59
0.907
0.0254
3.048
1054.88
2.326
429.907
0.06309
Openings
in.
0.0469
0.0234
0.0117
0.0059
0.0029
0.0021
0.0017
To Obtain
SI Unit
gram
megagram (Mg)
metric ton
meter (m)
meter (m)
Joule (J)
J/g
ng/J
liter/second
(1/s)
mm
1.18
0.60
0.30
0.15
0.075
0.053
0.045

-------
                                SECTION 1
                               INTRODUCTION

     Coal sampling, whether performed manually or automatically, must
extract a quantity of coal much smaller than the original lot for
laboratory analysis.  The sample, to be representative, must have the
same characteristic qualities and constituents as the entire coal lot.
However, Inherent variability 1n the coal parameters and the Imprecision
of the entire measurement process make sampling a rather technical
subject.  This report addresses the methodology, equipment, and modeling
of the coal sampling and analysis procedures.
     Section One of the report provides an Introduction concerning the
subject of the report and the contents of each section.  Conclusions are
provided In Section Two.
     Section Three discusses the various sampling techniques and
equipment.  The most widely regarded standards for sample collection and
preparation were established by the American Society for Testing and
Materials (ASTM).  Their Method 0 2234 details the minimum number and
weight of the Increments and the size of the gross sample needed to
provide a stated level of precision.  ASTM also evaluates the conditions
under which an Increment 1s collected.
     The collection of coal samples by manual and mechanical means 1s
next reviewed.  Most mechanical systems are based on the ASTM standards
for cutter speed and size.  The specifications and operation of several
augers (for manual sampling) and cutters (for mechanical sampling) are
given.
     Section Four discusses the Industry's sample preparation and
analytic techniques that are currently 1n use.  The primary sample may

-------
weigh as much as several hundred pounds whereas the laboratory requires
but 50 to 100 grams.* Hence sample preparation techniques such as
a1r-dry1ng, crushing, sample division, pulverizing, and mixing are
employed.  Procedures and equipment to perform each of these steps are
described with particular emphasis on Identifying factors which may
Increase the variance or reduce the representativeness of the resulting
sample.
     Next, the analytical methods used to determine moisture, total
sulfur, ash content, and calorific value (heat content) are described.
The appropriate ASTM and International Organization for Standardization
(ISO) methods are given and equipment available to perform the analyses
are described.  Quality control measures to ensure the reliability of the
analytic methods and the precision of the data are suggested.
     Section Five discusses the potential for sulfur loss as the coal
moves from bunkers through the pulverizers and the boiler.  Nuggets of
pyrltlc sulfur may be removed by the pulverizers.  Mass balance
relationships are difficult to perform; they do suggest a sulfur
retention 1n the bottom ash and fly ash as well as some S03 mixed 1n
with the SO- 1n the stack emissions.  Several sulfur retention
guidelines for EPA's consideration are offered.
     Section Six Identifies common coal sampling and analysis techniques
used by coal-fired utilities to evaluate their fuel.  This Information 1s
Important to the study because 1t Indicates how closely current Industry
practices match the proposed EPA Reference Method 19A requirements.
     Section Seven discusses 1n a relatively non-technical manner the
various coal sulfur parameters.  Dally SO  stack emissions are
estimated on the basis of dally composited samples of coal.  The accuracy
and precision of these estimates 1s of vital concern to EPA.  This
section discusses the role of time series data analysis, variances, and
auto-correlation on the estimation of current and the prediction of
*To convert from English units to metric units,  see the conversion table
 on Page 155.

-------
future emissions.  The role played by the measuring process will be
explicated.  The Importance of the various coal sulfur parameters .
v1s-a-v1s compliance with emission standards 1s described.  Techniques
which may be used to reduce variability and/or correlation are Included.
     Section Eight describes a theoretical statistical model for
estimating and predicting emissions data from coal sampling and analysis
data.  It 1s assumed that as-bunkered coal can be represented as Q
discrete packets of homogeneous material per day.  The computer
Implementation 1s based on Q = 1,008,000.  The cutter passing through the
stream collects L of these packets per Increment (computer Implementation
has L = 24).  The dally composite C 1s made up of N Increments (ASTM
recommends N = 35 for some boilers).
     The actual data X 1s made up of  the true value C plus a possible
measurement error.   The model  assumes that X can be described using a
fixed time series model.   Estimation  for the parameters of this model are
provided.
     Section Nine considers the measuring error 1n detail.  The data X
will be more representative of the actual coal C 1f the measurement error
1n the sampling and analysis procedures 1s minimized.  Considerations of
sampling error, sampling location, sample lot size, sampling system
performance, sample preparation, and  sample analysis are discussed; some
elements of a quality control  program to minimize these terms are given.
     Section Ten discusses bias.  Systematic error, or bias, 1n a
sampling system will consistently misrepresent the actual coal 1n the
same way.  This can lead  to Inaccurate (and probably misleading)
analytical results.  To prevent this, a sampling system can be bias
tested.  The proposed ASTM stopped belt procedure and one based on the
time series model presented earlier 1n Section Eight are offered as
possible bias tests.  Using the known size-frequency bias for coal from
the EPRI Coal Cleaning Test Facility  and the known sulfur size-frequency

-------
distribution for two different coals, the simulated sulfur bias was found
to be negligible.
     Tests of the model are presented 1n Section Eleven.  The statistical
model was verified by fitting 53 data sets.   Though the model 1s simple
and general purpose, there was no real Indication of lack of fit.
Splitting the data Into pieces produced consistent estimates for the
model parameters.  Long term statlonarlty (I.e., mean, variance, etc.,
constant over time) does not appear to hold.
     Last, Section Twelve, describes the beginning of a causal model
between emission data and coal and boiler characteristics.
     It 1s our conclusion that additional work will need to be done to
refine the models 1n order to obtain, better  fits to actual data.  This
report represents only a first approximation to the true situation and
should be read 1n that light.

-------
                                SECTION  2
                     CONCLUSIONS AND  RECOMMENDATIONS

     Coal  sampling and  analysis  (CSA)  procedures  cannot guarantee
"correct"  results.  Inherent coal variability and the representativeness
of the sampling and preparation  procedures lead us to conclude that the
resulting  50-100 grams  of coal  per day analyzed 1n the laboratory may not
have coal  parameters exactly equal to the  dally average of the coal.  The
laboratory results rely on fallble humans  which may Introduce additional
Inaccuracies.  Thus, any discussion of CSA must address the statistical
Issues of  estimating the true emissions  from the scattered sulfur data.
     This  study presents a theoretical model for coal sulfur data
Involving  time series modeling.   Extensive verification of the model
using real coal data does not suggest any  widespread  lack of fit.  On the
contrary,  the model seems to do  a very credible Job in fitting the data.
The model  can be used by EPA to  evaluate the Impact of various stack
emissions  standards and/or averaging periods on the mean level of
compliance coal.  Alternately,  1t can be used by EPA to offer guidelines
to the Industry as to how 1t could meet proposed stack emission standards.

-------
                                SECTION  3
              ASSESSMENT AND STATISTICAL EVALUATION OF  COAL
                    SAMPLING TECHNIQUES  AND EQUIPMENT
    The purpose of this section 1s to document Industry accepted coal
sampling methods and define the specific sampling principles Involved.
The objectives of coal sampling are discussed, Including the procedures
and rationale for collecting a representative sample.  The development of
accepted coal sampling guidelines are presented.  Sampling equipment 1s
evaluated according to Its effectiveness 1n obtaining a representative
sample according to accepted standard methods.  A discussion of sampling
considerations and objectives Include suggested procedures to achieve the
level .of accuracy needed to meet specific standards.  Coal sampling 1s
presented as a means of predicting SO- emission levels.

3.1  OBJECTIVES OF COAL SAMPLING
    The sampling of coal, whether performed manually or automatically,
must extract a quantity of coal much smaller than the original lot but
with proportionately the same characteristic qualities and quantities
present 1n the entire lot.  It has long been realized that the properties
1n coal are not distributed uniformly.  The variability of coal makes 1t
difficult to collect a sample that 1s representative of a large mass of
coal.  For Instance, grab samples of coal from the same source may show
different analytical values 1f tested 1n different laboratories or by
different technicians 1n the same laboratory.

-------
    Besides coal's Inherent variability,  other factors  such as how the
coal was handled,  or how the samples were obtained,  affect the collection
of a representative sample.  The coal may have become segregated during
loading, transport, or unloading operations so that  the particles are
grouped together by size.  When samples are taken from a stationary
source, such as a  coal storage pile or railroad car, 1t 1s difficult to
obtain an accurate sample because the material located  1n the center of
the pile will be Inaccessible when conventional sampling techniques are
used.  Generally,  samples taken around the pile will be limited by the
depth of penetration of the sampling device Into the stationary source.
Similarly, when samples are taken from a moving stream, such as a
conveyor, they should be taken from the entire width of the belt to avoid
biasing the sample.
    To be effective, a sampling plan must employ measures to reduce the
effect of segregated particles, minimize the effect  of the variability of
the coal properties, and Identify any mechanical' bias due to the sampling
method.  Sampling material from a conveyor or a chute through which the
coal 1s flowing provides access to a cross-section cut of the entire
stream.  This cross-sectional cut will provide a characteristic sample
even though the vertical distribution of material on the conveyor may be
segregated by particle size.
    Mechanical sampling bias may be reduced by the use of automatic
equipment which 1s not dependent on human discretion for operation.
These systems are generally elaborate and have been  designed for a
specific plant's application.  Manual sampling methods can be used, but
care must be taken to ensure that the sampling technique has been
consistently applied.
    Sampling personnel must consider the variability between discrete
units of coal when attempting to collect a sample which 1s representative
of a specific lot.  As an example, a 10,000-ton lot  of coal may be made
up of 100 discrete units or railroad cars of 100 tons each.  If the

-------
variability between railroad cars 1s high because of differences 1n
loading procedures and coal  characteristics,  a composite made up of one
Increment collected from every tenth car may  be Insufficient to represent
the entire 10,000-ton lot.   In this case, Increasing the number of
Increments or Increasing the number of railroad cars sampled, will
produce a composite which better characterizes the entire lot.  On the
other hand, 1f the variability 1s expected to be relatively low, then
sampling from each car may be an extensive effort with little or no extra
benefit.
3.2  SAMPLING GUIDELINES
    With all these factors contributing to the Inaccuracy of coal
sampling, there has been concern over the reliability of coal analysis
data.  Inaccurate data, whether due to an error 1n sampling, poor
analytical techniques, or some other factor,  can result 1n data misuse.
Because of this, there are many opinions concerning guidelines for
sampling coal and the establishment of standard methods.
    As early as 1914, the United States Bureau of Mines (USBM) presented
a procedures, written by A.  C. Meldner, for  sampling coal 1n a
mine.   .  The Bureau also developed guidelines for representative
                     (3)
sampling for analysis*  .  The author, G. S.  Pope, points out that 1t
1s difficult to obtain a representative sample from a stationary source
such as a rallcar or storage pile.  He further states that the only
representative sample that can be collected from a heterogeneous mixture
of coal 1s one from a moving stream of coal as 1t 1s being loaded or
unloaded.  Pope also offers  general directions for the number, size, and
frequency of sample Increments.
    Pope states that the conclusion reached by tests run by the USBM
Indicates that the gross sample should not be less than 1,000 pounds.
The Increments should be systematically spaced so that the entire

-------
quantity 1s represented 1n the gross sample.   Pope suggests Increments of
10 to 30 pounds, but he points out that Increment size 1s governed by the
size and weight of the largest pieces of coal; Increments of greater than
30 pounds may be necessary.
    A more specific guideline was provided by Bertrand A. Landry 1n
another USBM publication^  .   The author applies the theory of random
sampling to coal sampling methods.  Specifically, Landry used random
sampling theories to determine the optimum number of Increments to be
collected to yield a gross sample of definable quality.  The
determination of the number of Increments 1s based on the desired
representativeness of the gross sample and assumes that the variability
of the Increments 1s known.  Landry presents an equation for this
determination:
                                       2
                                N =  SVI
                                      2
                                     S(N)W

which states that N, the sufficient number of Increments, can be found by
                                                         2
dividing the variability of the Increments of weight W, S.., by the
variability of gross samples of accepted representativeness,
S,NvU.  This 1s based on the premise of the random chance of
 \N)\H
choosing each Increment, and on all Increments being the same weight.
The total weight of coal being sampled 1s not considered 1n determining N
for random sampling.  The weight of each Increment 1s assumed to be
constant, but no specific weight 1s suggested.  Landry then determines
that systematic sampling will produce a sample as representative as
random sampling when the total lot size 1s relatively large.
    The U.S. Steel Corporation has also written directions for providing
                                   (4\
representative samples for analysis^  .  They advise removing a full
cross sectional cut from a moving stream or stopped belt conveyor, but
they discourage sampling from railroad cars and other stationary

-------
sources.  U.S. Steel recommends the collection of one Increment every 30
minutes for moving streams and provides guidelines for gross sample
requirements using different sampling schemes.  A minimum of 1,000 pounds
1s specified for a gross sample.  A sample Increment weight of six to
eight pounds 1s recommended when samples are collected from chutes or
stationary sources.  When sampling from a stopped belt a much larger
Increment 1s required.  The cross-sectional cut should be a minimum of
one foot 1n length.
    To achieve a more uniform, systematic method for collecting coal
samples, the USBM published a paper for government employees to use as a
guideline 1n obtaining representative samples^  .  This paper suggests
systematic collection of a large number of Increments of equal weight.
The exact number and size of Increments 1s dependent on the maximum size
of the coal.  In the case of a coal whose top size 1s 3/4 Inches, 50
Increments of 10 pounds each are required.  For larger sized coal, the
number.of Increments remains at 50, but the Increment weight 1s
Increased.  For example, a coal whose top size 1s 8 Inches requires
Increment weights of 20 pounds each.  The USBM recommends a full
cross-stream cut be taken from a moving stream.  They discourage top
sampling from railroad cars or barges.  If, as a last resort, top
sampling 1s used, Increments should be taken from throughout the car or
barge.
    All of these references give general recommendations for collecting a
gross sample of coal, but none 1s an Industry accepted standard method.
The most widely regarded standards were established by the American
Society for Testing and Materials (ASTM)(6).  The ASTM Issued
recommended procedures for a variety of sampling situations.  The ASTM
Method 0 2234 details the minimum number and weight of Increments and the
amount of the gross sample needed to provide a stated level of precision.
                                     10

-------
    The ASTM method of determining the number and weight of increments,
                          i K\
similar to USBM guidelines  '.   The number of increments needed to
achieve a required degree of precision is calculated using a relationship
between several sampling variances and the increment number such that:
where:
    N = number of increments,
    W = weight of each increment,
     2
    S^ = segregation variance,
     2
    S  = random variance of a 1  pound increment,
     2
    Sg = variance of 1 gross sample,
     2
    S.  = variance of division  and analysis,  and
    P = number of analysis samples.
The ASTM method used assumptions from a specific sampling situation to
calculate that a minimum of 15 increments was necessary for each gross
sample representative of a lot of 1,000 tons of mechanically cleaned
coal.  In the situation where the lot size is expected to be more than
1,000 tons, one gross sample can be collected to represent the higher
tonnage.  However, more increments are required.  The number of
increments can be expressed as a ratio of the expected lot size to the
1,000 ton lot size, as follows:
                     N2 = N! , /total  lot size (tons)
                            V1,000 tons
where:
       = 15 increments
       = number of increments required.
                                     11

-------
    The weight of each Increment 1s based on top size to minimize bias
due to particle size.   Generally, the larger the particle size present 1n
the lot of coal, the heavier the Increment should be.  For a top size of
5/8 Inch, ASTM specifies 2-pound Increments.  A top size of 2 Inches
requires 6-pound Increments, while a 6-1nch coal requires 15-pound
Increments.
    The ASTM also evaluates the conditions under which an Increment Is
collected, such as: stopped-belt cut, full-stream cut, part-stream cut,
and stationary sampling.  ASTM points out that an automatic or mechanical
sampling method, without human discretion, 1s more precise than manual
sampling.  They classify each condition under which an Increment 1s
collected.  The highest classification, or most precise method, 1s the
stopped-belt cut removed by a mechanical cutter with Increments spaced
systematically.  The stopped-belt technique allows all of the particles
1n a cross-section of the belt to be collected,.eliminating segregation
due to.particle size.   ASTM specifications restr-lct the cutter speed to
18 1n./sec and Indicates the minimum cutter width (size of the opening)
of 2 1/2 to 3 times the topslze of the coal.
    ASTM also considered the ratio of maximum particle size and the large
variation of ash content among coal pieces of different size.  Ash .
content was used because 1t was believed to be the most sensitive measure
of variations 1n coal  quality.  ASTM specifies that no less than the
minimum Increment weight be collected so as to reduce the bias due to
particle size and ash content.
    These factors were discussed by Ornlng, who studied the distributions
of particle sizes, ash contents, and their effect on the precision of
sampling1  .  Ornlng estimated the Influence of various sizes and ash
contents of coal on the variability of ash content 1n duplicate coal
samples.  He found that the sampling variance 1s more dependent on
particle size and the possible rejection of large pieces of coal than on
Increment weight.  Ornlng would permit a lesser number of smaller
Increments when coal of small top size 1s being sampled.
                                    12

-------
    The theory of sampling based on the ASTM sampling principles was also
                               / g\
discussed by Vlsman and Aresco/ '   The theory provides estimates of
the variance of a coal, then provides the minimum number and weight of
Increments needed to attain a desired level  of precision.
    Another view 1s taken by Bertholf  '  who theorizes that a large
number of almost weightless Increments should be taken.   The minimum
Increment weight proposed by Bertholf's theories 1s Impractical based on
the mechanical limitations of samplers.  Therefore, he proposes the
collection of a primary Increment, larger than needed, which 1s then
resampled, resulting 1n a secondary Increment closer to the minimum
weight requirements.  This method would apply especially to mechanical
coal samplers which take primary Increments large enough to meet ASTM
requirements each time the device 1s activated.  The quantity of each
primary Increment collected 1s a function of the amount of material 1n
the moving stream and can exceed several  hundred pounds when these ASTM
specifications are observed.  This bulk of material far exceeds the
quantity needed for analytical purposes and Is much larger than the ASTM
minimum Increment weight of 6 pounds for 2-1nch topslze material.  In a
study of mechanical coal sampling systems, Bertholf showed that large
representative Increments could be collected and reduced to smaller
samples without a loss of precision^   .
    Coryell et al. tested a different type of mechanical sampler which
took primary Increments of approximately 50 pounds and reduced each to a
secondary Increment of 3 pounds.  These tests showed that the desired
ASTM precision could be achieved by Increasing the size of the primary
Increments while decreasing the number of Increments^   .
    It 1s generally accepted within these various guidelines that the
number of sample Increments should be based on lot size and sampling
variance with Increment weight being determined by considering maximum
particle size.  However, a different approach 1s suggested by
  n 2 13}
Syv <•• -/   Sy Dases h^s methods on the theory that coal
                                 13

-------
characteristics are not randomly distributed but are autocorrelated.
Using the autocorrelation theories,  Gy estimates a sample size based on
particle size and shape.   The minimum number of Increments would then be
determined from the estimated total  sample size, the flowrate of the
stream, and the width and velocity of the cutler.  Other guidelines
estimate a total sample size based on lot sizes, and the minimum number
of Increments 1s based on variance and the required precision.  The
theory that coal 1s autocorrelated and not randomly distributed 1s also
                             (141
addressed by Gould and V1smanv   .
    Another recommendation that Gy makes concerns cutter width and
speed.  Gy conducted experiments to determine the critical value of these
two parameters.  The critical value for the cutter width was determined
to be three times the particle diameter, similar to the ASTH
requirements.  The critical value of the cutter speed was determined to
be .6 m/s, a value slightly higher than the ASTM recommendation.
    Table 1 summarizes the sampling guidelines and methods presented 1n
this section.
3.3  SAMPLING EQUIPMENT
3.3.1  Manual Sampling
    Although the recommended procedure for collecting a coal sample 1s
with an automatic sampling system, manual sampling 1s sometimes the only
alternative.  Manual samples can be taken from a stationary source or,
more reliably, from a moving stream.  In this discussion of coal sampling
equipment, some commonly used hardware will be described.
    In stationary source sampling, such as from railroad cars or storage
piles, shovels, buckets,  probes, augers, and other Instruments can be
used.  The simplest, and perhaps most Inaccurate means would be sampling
with shovels or scoops from a stationary source.  Although discouraged,
there are several guidelines for this type of sampling (2,4,5).  All of
them recommend taking several Increments from different points and at a
                                   14

-------
                   TABLE 1.  SUMMARY OF SAMPLING GUIDELINES

Author
USBM/G.S. Pope


USBM/B.A. Landry


U.S. Steel


USBM/N.H. Snyder


ASTM


A. A. Or ning
W. M. Bertholf


Dr. P. Gy

Recommended
Method
From a moving
stream

From a moving
stream

Full cross-
section.
stopped belt
Full cross-
section.
moving stream
Full cross-
section,
stopped belt
Not specified
Not specified


Mechanical
method
Number of
Increments
Spaced
systematically
over entire lot
N= S2 y
s* (N)U

One every
30 minutes

50


Minimum of 15


Number and size of
Minimum number
based on coal's
variability
40 to 50. based
on estimated
Size of
Increments
10-30 pounds


All increments
must be same
size
6-8 pounds


10-30 pounds

,
2-15 pounds


increments depends
Not specified


Dependent on
speed and width
Gross
Sample
1,000 pounds


Not considered


1,000 pounds


500-1,500 pounds


Dependent on
increment number
and size
on variability of
Not specified


Dependent on size
and shape of

Recommendations
Larger increments result in
more accurate sample

Sampling is based on
variability of coal and
desired precision
Orderly spacing of increments
to represent entire lot

Increment size dependent
on coal top size

Increment size dependent
on coal top size

coal's size and properties.
Proposes collection of a
primary increment, which
is then resampled
Recommends a mechanical
sampler with a large
sample size        of cutter          particles               cutter width

-------
consistent depth.   This method win  not always  produce a  representative
sample of the entire lot because only the uppermost particles  1n the
storage pile or transport vehicle are accessible to the sampling device.
    An auger can be used to obtain a core sample of consistent depth Into
the pile.  As an example, the Holmes Model  13 has a 30-1nch auger (F1g
3-1) with a 4-1nch diameter which limits 1t to  coal with  a topslze of
1.25 Inches or less.  Larger truck-mounted, boom-operated power augers
are available and  are commonly used to collect  feedstock  samples from
railroad cars.  The augerlng technique may not  accurately represent fine
particles because  they tend to fall  out of the  sampler prior to discharge
Into the sample collection container.  Although the use of an  auger
allows deeper penetration Into a pile than shoveling,  quite often the
Inner portions of  the pile remain unsampled.
    Prelser/HINECO offers slotted probes of varying lengths and diameters
(F1g 3-2).  These  devices may be Inserted Into  a pile  but are  limited to
materials with a nominal topslze of less than 0.5 Inch.
    Samples taken  from a moving stream by manual techniques are more
precise than samples from a stationary source,  but they are subject to
human error.  The  human factor Involved 1n accurately  repeating the
cutting process adds to the sampling error.  For sampling material from a
falling stream, the simplest devices Include shovels,  scoops,  and
buckets.  The shovel or similar device should have raised sides and
capacity sufficient to hold a cross-sectional cut without overflowing.
Using the ASTM guidelines of sampling at a rate of 18  1n./sec  or less and
using a cutter which has an opening at least 2.5 to 3  times the topslze
of the coal, collecting a sample from conveyor moving  coal at a rate of
500 tons/hr. would yield a sample weighing 200 to 300  pounds.
    An Increment this size would be very difficult to  handle manually;
therefore, a partial cross-stream cut should be considered.  Although
less precise then the full cross-stream procedure, multiple partial
cross-sectional cuts should be taken at different locations until the
                                    16

-------
Figure 3-1.   Holmes sample probe, Auger style.
Figure 3-2.  Sample probe, aluminum, 66" long,
           11 openings, 1 3/8" O.D.
      Figure 3-3.  Preiser hand sampler.
                      17

-------
entire stream has been sampled.   Extreme care must be used when taking
manual samples are taken from a  moving stream, because the collection
container generally fills very rapidly and may be quickly swept from the
operator's hands.  Samplers should use a device that has a secure handle
and that can be rested on a solid object such as a chute access door
frame or a railing along a conveyor transfer point.
    Prelser/HINECO distributes a hand sampler which 1s a good tool for
obtaining a cut or Increment from a falling stream (F1g 3-3).  The pan or
trough section comes 1n varying  sizes as large as 8 Inches wide, 8 Inches
deep, and 48 Inches long.  Depending upon the flowrate of the stream and
the topslze of the particles, a  full cross-sectional cut can be taken
with the larger samplers that meets the ASTM cutter opening size
specifications.  As with all manual sampling devices, human discretion 1s
Involved, which can affect the precision between successive Increments.
    Manual sampling from a moving belt may, at some facilities, pose a
serious safety problem.  If plant operations allow, the conveyor should
be stopped and a complete cross-section taken.  The width of the cut
should be at least 2.5 to 3 times the size of the largest particle.  The
sides of the cut should be parallel.  Care must be taken to remove all
particles within the measured cross-section.  A hoe or similar tool can
be used to remove the material along with a brush for the finer
particles.  This stop belt method 1s the basis by which ASTM evaluates
all automatic and manual methods for sampling of moving streams for bias.
    Manual sampling techniques,  although subject to errors associated
with human discretion, may be used to effectively collect samples of
definable quality.  Automatic sampling systems, once designed, Installed,
and tested for a specific plant application, however, will routinely
produce samples which are cost-effective and, 1n terms of quality,
generally exceed those collected by manual methods.
                                    18

-------
3.3.2  Mechanical Sampling
    The collection of coal samples by mechanical  means 1s considered a
much more precise and representative method than  manual sampling.  The
ASTM bases this assumption on the fact that an automatic sampler will not
select an Increment on a discretionary basis.  There 1s no human element
Involved.
    Most manufacturers of mechanical sampling systems have used designs
based on ASTM standards   '    .   Several types are now 1n use Including
full cross-stream cutters, rotating scoops, augers, and rotary arm
samplers.  The reliability and accuracy of different automatic systems 1s
difficult to quantify and 1s a function of the Installation of the
equipment, the physical characteristics of the material to be sampled,
and the operating techniques used by the facility personnel.  As
previously mentioned, systems of this type as generally evaluated based
upon a comparison with the ASTM stopped belt technique (ASTM D-2234-76,
Appendix A).  A standardized bias test does not exist, and 1n many cases
sampling systems manufacturers have developed their own tests to evaluate
the performance at the time the system 1s Installed.  A procedure
proposed by Ornljig' ' consisted of compositing replicate gross samples
from Interleaved Increments and determining the variance.  This procedure
was employed by Aresco and Ornlng^  ' 1n a U.S. Bureau of Mines study.
    The cross-stream cutter type system 1s widely accepted as providing a
representative primary sample Increment.  The Ramsey/Gal1gher Engineering
Company designed a cutter sampler to meet ASTM specification 0-2234-76
(Figure 3-4).  A full cross-stream cut 1s taken as the coal discharges
from a conveyor head pulley.  The cutter starts at one side of the belt
discharge and cuts through to the other side so that the entire
cross-section 1s represented.  The cutter speed and opening can be
adjusted to meet ASTM requirements.  The Ramsey/Gal1gher sample cutters
are designed for flow rates of up to 10,000 tph.
                                    19

-------
              CONDENSED SPECIFICATIONS
             Model B-1800 and  C-1800 Primary Sampling
             Machine:
             Capacity:
             Particle Size:
             Cutter Stroke:
             Cutter Type:
             Drive Speed:
to 10,000 tph
up to 12" top size
72" to 96" or as req'd.
"C" 48'-96"
Hydraulic: to 30 in./sec.
Electric Chain Drive: 6, 12 or
18 in./sec.
Figure 3-4.   Ramsey/Galigher primary sampling machine.
                             20

-------
    Similar systems are produced by the Denver Equipment Division (Figure
3-5).  The Denver model samples a stream as  1t falls  through a chute.
The cutter speed and opening are adjustable,, and  a  full  cross-sectional
cut 1s obtained.
    The Quality Control Equipment Company also has  a  cross-cut sampler
designed for chutes.  However,  the maximum particle size 1s set at two
Inches (Figure 3-6).
    The McNally Belt Sampler takes the discharge  from a  belt and passes
1t through a chute section where the sample cutter  removes a full
cross-section (Figure 3-7).  The McNally design 1s  consistent with ASTM
specifications.  The cutter opening can be set to three  times the size of
the largest particle.  For higher tonnages,  McNally designed a travelling
plate sampler (Figure 3-8).  The cutter 1s driven through the stream of
coal falling through a chute.  The plate-sampler  1s suitable for large
size material, and the speed of the cut can be adapted to ASTM
specifications.
    A traversing sampler which obtains a full cross-sectional cut from
coal flowing 1n a chute has been developed by Harrison R. Cooper Systems,
Inc. (Figure 3-9).  The sampler maintains a constant cutter speed of 18
1n./sec and can handle very large loads.  The same  company provides a
different method of taking a full cross-section cut.   The rotatlng-scoop
sampler collects the sample from the entire stream at the discharge end
of a conveyor belt (Figure 3-10).  The scoop does not cut through the
width of the stream as the cutter samplers do, but  swings directly Into
the entire outfall head-on.
    The Bristol Engineering Company has designed  a  rotary arm scoop
sampler which takes a sample directly from a moving belt (Figure 3-11).
The scoop 1s situated above the belt, then swings down,  passing through
the coal while moving 1n the same direction but slightly faster than the
belt.  At the top of Its arc, the scoop 1s stopped  and empties Into a
                                      21

-------
                  Model  Feed    Feed
                  Travel  Rate    Opening  Cutter
                  (inches) (TPH)    Size     Opening
                  12-FI   0-25    6" x 6"   1" - 3"
                  18-FI   25-100   10" x 10" 1" - 3"
                  30-FI   100-250  10" x 18" 4" - 10"
                  24-FI-  0-25    10"x10" 1" - 3"
Figure  3-5.   Denver  Equipment Division's chute  sampler,
                              22

-------
                               I>MM •» *ra-
              TtC* »».—.—•
                     \
1
ii
Figure 3-6.  Quality Control Bguipnent Company's cross cutting  sampler.
                   Figure  3-7.   McNally belt sampler.
                                 23

-------
Figure 3-8.  McNally plate sampler.
              24

-------
Figure 3-9.   Harrison R.  Cooper Systems,  Inc.  traversing cutter.
                                25

-------
Figure 3-10.   Harrison R.  Cooper System's rotary scoop sampler.
                             26

-------
^
X
±,
^r
«. .
:». /
l*'A?W
i%?iK/
Vf.ZSssI

+~—c~




—
^.
\
%
t
1
1
% "/ .
V ..^rxsssla^
Figure 3-11.   Bristol Engineering Company's rotary arm  scoop.
                             27

-------
hopper.   This system takes only a 3-1nch wide sample from the flow, a
partial  cross-stream cut.   The validity of this  sample 1s dependent on a
homogeneous, non-stratified flow of material  on  the belt.  The collection
of a sample by the rotary  arm scoop may also  be  subject to some
segregation due to size, because 1t takes the sample from the same
section of belt each time.
    The many different approaches to mechanical  sampling allow for many
different applications and consideration of specific requirements.
Although mechanical sampling 1s more precise  than manual, not all
mechanical methods collect equally representative samples.  The design
should Include a means for taking a sample from  the entire cross-section
of the coal stream.  Other Important factors  to  consider are the speed of
the cut and the size of the cutter opening.
    In general, the cutter should move at a uniform speed to ensure that
the entire cross-section 1s represented 1n equal proportions.  The speed
should also be slow enough1 to prevent segregation and rejection of
particles due to disturbances of the coal stream.  The cutter opening
should be large enough, at least three times  the size of the larger coal
particles, to allow equal  representation of all  size particles.
    The preferred method of sampling would be the method that obtained
the most precise, representative Increment.  Table 2 shows the order of
preference for each sampling scheme discussed 1n this section.

3.4  SAMPLING CONSIDERATIONS
    It 1s useful when considering a sampling system to have as a goal a
desired end result.  The goal may be the collection of a truly
representative sample of the coal entering a boiler, perhaps to assess
the efficiency of the boiler or to aid 1n designing ash handling
systems.  Coal sampling 1s also used to determine a coal's
characteristics, and subsequent market price.  Coal sampling and analysis
1s now being used as a continuous SO  emissions  monitoring method.
                                    28

-------
 TABLE 2.  ORDER OF PREFERENCE OF SAMPLING PROCEDURES AND METHODS
Collection Procedure    Collection Method
                          Order of Preference
Stopped belt cut

Stopped belt cut
Full stream cut
Full stream cut
Part stream cut
Part stream cut
Stationary sampling
Mechanical, automatic
system
Manual sampling
Mechanical, automatic
system
Manual sampling
Mechanical, automatic
system
Manual sampling
Mechanical, automatic
system
2
3

4
5

6
7
Stationary sampling     Manual sampling
                                29

-------
    In general,  1f the amount of sulfur present 1n the coal  going Into
the boiler 1s known,  then the SO. emission level  can be determined by
quantifying the  amount of sulfur lost 1n combustion and trapped by air
pollution control devices.  With emissions monitoring as a goal, several
factors must be  considered when Implementing a coal sampling plan.  The
first consideration 1s the accuracy of the sampling procedures.  A choice
must be made concerning manual versus automatic sampling and a stopped
belt technique versus a moving full-stream cut.  The number, frequency,
and size of Increments are also Important considerations.  Other
Important factors are the behavior of the coal 1n the boiler, and the
amount of sulfur 1s removed by pollution control  devices such as
electrostatic predpltators and flue gas desulfurlzatlon systems.
    The accuracy of sampling systems should be the most Important
characteristic considered when designing a sampling protocol.  An
automatic sampling system utilizing the stopped belt technique 1s the
preferred system because 1t 1s the most precise.-  The number, frequency,
and size of Increments should be based on a statistical determination of
the minimum requirements for a representative sample of that specific
type of coal.  The ASTM standard 0 2234 1s one such accepted statistical
guide.  The number, frequency, and size of Increments 1s a very Important
factor to consider especially 1n cases of coal with high sulfur
variability.
    The accuracy of a sampling procedure can be determined by running a
bias test.  A bias test 1s a statistical comparison of samples obtained
from the plant's routine sampling plan and samples obtained from a
stopped belt technique.  In the case of a mechanical sampling system, the
sampler would be run for a set length of time and the final samples
retained for analysis.  Periodically during the run of the sampler, the
belt would be stopped and a full cross section of the coal removed and
retained for analysis.  The length of the cut should be at least 20 times
the top size of the coal.  The stopped belt sample can be divided using a
                                    30

-------
riffle or mechanical divider.  Each of the two sets of samples should be
analyzed for the parameter for which the variance calculations are to be
made.  In the case of SO- emissions monitoring, analysis for total
sulfur content should be made.  A statistical comparison of the results
1s then used to determine the level of variance between the normal
sampling point and the bias test sampling point.  If the bias 1s within
acceptable limits, then the sampling system 1s considered free of bias.
    Another Important factor that must be considered when using coal
sampling and analysis as an SO. emissions monitoring method 1s the
correlation of the sulfur 1n the coal to the SO  emissions.  First the
time lapse between sampling of the coal and burn of the lot of coal that
the sample represents must be determined.  If the coal 1s sampled from a
belt which feeds the coal storage bunkers, the retention times of the
bunkers should be known.  An 8-hour retention time would mean that the
coal burn and resulting SO. emissions would correspond to the coal
sampled 8 hours beforehand.
    The amount of sulfur reduction resulting from the combustion 1n the
boiler, and from air pollution control devices must be considered.  The
boiler efficiency, and efficiency of the sulfur removal processes such as
flue gas desulfuMzatlon and electrostatic precipitation can be used to
calculate the amount of sulfur going out the stack.
                                   31

-------
                                SECTION  4

             ASSESSMENT AND STATISTICAL  EVALUATION OF  SAMPLE
                    PREPARATION AND ANALYTICAL METHODS
    In this section, the Industry accepted procedures for preparing a
coal sample for analysis and analytical methods currently 1n use are
discussed.  Sample preparation techniques such as a1r-dry1ng, crushing,
sample division, pulverizing, and mixing are presented, along with the
equipment necessary to perform each step.  Factors that create a sample
bias or Increase the variance are Identified.
    The analytical methods used to determine moisture, total sulfur, ash
content, and calorific value of the samples are described, Including the
equipment needed to perform these analyses.  Quality control measures to
ensure the reliability of the analytical method and the precision of the
data obtained are suggested.
4.1  SAMPLE PREPARATION
    The collection of a representative gross sample using manual or
automatic methods often yields a bulk quantity of material which may
weigh as much as several hundred pounds.  Normally, 50 to 100 grams of
material are necessary to meet the analytical requirements of most coal
characterization tests.  The method used to reduce the gross sample to
the analytical sample must maintain the Integrity and representativeness
of the sample while meeting the particle size and weight requirements as
specified by the particular analytical methods.
    The preparation of a gross sample for coal analysis requires several
processing steps.  A1r drying 1s used to bring the moisture content of
the sample to equilibrium with the air 1n the room where further prep-
aration will take place.  Samples are reduced or crushed from a nominal
                                    32

-------
3-1nch top size to minus No.  4 or No.  8 mesh  used  to optimize the
reprodudblHty of subsequent processing procedures.  Sample division or
splitting results 1n a smaller quantity of  material  without a loss  1n the
sample's representativeness.   Another  sample  reduction process,
pulverizing to minus 60 mesh, 1s generally  required  for specific
analytical procedures.  Thoroughly mixing and homogenizing the analytical
sample 1s necessary so that the small  allquots which are taken for
Individual tests are representative of the  sample.
4.1.1  A1r Drying
    A1r drying of a gross sample 1s generally the  first step 1n  sample
preparation.  The purpose of  air drying coal  1s to bring Its moisture
content to equilibrium with the air 1n the  room where 1t 1s to be reduced
and divided..  The gross sample can be  air dried either 1n an air  drying
oven or on a drying floor.  The temperature 1n an  air drying oven should
be 10°C to 15°C above room temperature with a maximum oven temperature of
40°C to prevent rapid oxidation of the coal.   The  coal 1s spread  thinly
on trays 1n the oven and exposed to a  flow  of slightly heated air (Figure
4-1).  According to the ASTM 02013-72(1978) procedure(s), air drying
should stop when the loss 1n  weight of the  gross sample 1s less  tharr 0.1
percent/hour.  A1r drying on  a drying  floor 1s a much longer process,
because there 1s no heat source such as an  oven.  The sample must meet
the same drying time requirements as those  described when using  an  air
drying oven.
    To minimize moisture variability between  samples, several precautions
are necessary.  A1r drying by oven must not be done for an excessive
amount of time.  Oxidation of the coal must be avoided.  A1r drying on a
drying floor must be done 1n a room free from dust and excessive wind
currents.  These precautions  will minimize  the variance 1n this  stage of
sample preparation.
                                       33

-------
                              1/2 in. (1.27cm)
         TOP           y   Celotei  Insulating Board
                         -— 4-500 Wall Strip Heaters
                  -• Blower


_    34 in. (86.3 cm)      --
                                             l/2in. (1.37cm)
                                             Celotex Insulating
o o o o  o
O 0 O O  0
o o o  o  o
O 0 0 O  0
o o o o  o
             o o o o  o

              Shelf
             0 0 O 0 O
             o o o o o
             o o o o o
             o o o o o
                            Sin. (20.3cm)



                            1?

                            7 **


                            I
                           o
                           s
                                               4-l/2in m
                                               (II.4cm)
                                              ^-Air Inlet
                                               Holes

                                                 Air Outlet-f
                                                .ttoies..
     .36 in. (91.4cm)-

        FRONT
                        !              :    I
                       -   6in.(l5.2cni*.    -.— 24in

                                             SIDE
     Figure  4-1.   Air  drying  oven.
                        34

-------
4.1.2  Sample Reduction
    Sample reduction 1s the next step 1n sample preparation.   Reduction
1s the process where a sample 1s reduced 1n particle size,  without a
change 1n weight.  Crushers that are commonly used,  Include:  Jaw type
(Figure 4-2), hammer mill (Figure 4-3),  and rotary (Figure  4-4).
    The crusher should be large enough to process  the gross sample 1n a
reasonable amount of time.   A sieve test should be performed  periodically
on the product of the sample crusher, to assess the uniformity of
particle size.  Also, a uniform flow rate Into the crusher  should be
maintained to guard against segregation  by composition.   This may
Introduce a bias 1f, during periods of a low flow  rate to the crusher,
the more brittle particles  are selectively crushed first.  A  final
precaution 1s to make certain that no particles remain 1n the crusher
after this processing step  1s completed.  The material must be crushed to
a top size to pass through  a No. 4 (4.75 mm) or No.  8 (2.36 mm) sieve
based upon criteria established 1n ASTM 0-2013-72(1978).
4.1.3  Sample Division
    The next step 1n sample preparation  1s sample  division.  Here the
sample 1s reduced 1n weight without a change 1n particle  size.  Riffles
or sample splitters are used to divide the sample  stream Into two equal
parts (Figure 4-5).  One part 1s discarded, and the other remains as the
sample for further splitting or analysis.  It 1s Important  that the
riffle divisions are at least three times the size of the coal being
divided.  The coal should be fed Into the riffle 1n a slow, uniform
stream covering all of the  slots of the riffle.  The coal should not be
allowed to build up 1n or above the riffle slots,  and 1t  should flow
freely through the slots.  These precautions are necessary  to minimize
the variance due to riffle  division.
                                      35

-------
JawGvshers
it





I
CiMrifr 	 _^ ._._... MWffM
W ^") teftl/W/,* MQfMpOHW t*rtfl(f* WWtll Hftgtlt (Lni,t
tit v, i r-itr r-tr r••*• 4000
• jo in n TO S'4" f-r r-f 9500
9llS 1070 70 r<- S'-O- 4'-4- 17000
Figure 4-2.
                                 crusher.
                            36

-------
Hammermills
Size
OOSM
OOHP
OSM
09HP
ISM
OI8HP
2SM
OHHP
Capacity
Tons/Mr.'
l-IVi
2-4
2-9
2-9
S-20
3-12
12-70
S-IS
Horsepower
3-S
5-20
10-30
5-25
25-75
15-20
50-100
15-20
length
I'-l"
2'-5"
t--r
r-r
S'-r
3'-r
«'-!•
y-r
Width
2'-4-
3--0-
4'-l'
*-2-
s'-a-
f-n-
7'-0"
S'-S-
Height
J--5"
t'-l !"•
3--1-
2'-9~
3'-9"
2'-9-
*-5-
2'-9-
Weight

-------
NOM.
SIZE
6x8
ano
tone
DIMENSIONS
H
IZV
uW
l7'/«"
L
n"
13"
23"
W
II"
13V
I8*e
RIFFLES
Opening
Width
V
V
I"
NO.OI
Divisions
14
12
18
                  CLEANINO SHUSH FURNISHED
                                           SAMPLE PANS
Figure  4-5.   Riffle.
             38

-------
    Samples can also be reduced 1n bulk by  mechanical  sample dividers.
Although not all automatic sample dividers  split the sample 1n halves,
accurate division can be achieved by extracting and  compositing a large
number of Increments from the sample.   The  ASTM 02013-72(1979) procedure
specifies that at least 60 Increments  be taken.  One type of mechanical
sample divider which does halve the sample  1s  the rotary plate divider
(Figure 4-6).  The rotating plate directs the  sample by centrifugal  force
toward the outlet.
    Another type of mechanical divider 1s the  rotating cutter (Figure
4-7).  Ramsey/Gal1gher makes a rotary  sampling machine with a sample
cutter that revolves on a central shaft.  The  cutter opening and speed
can be set so that a specific percentage of the sample 1s divided and
remains as the sample for analysis.
    The Denver Equipment Division's Vezln sampler 1s another rotary
cutter sampler that can be used for sample  division  (Figure 4-8).  The
cutter turns at a constant .speed and splits a  fixed  percentage of the
sample.
    A rotary sampler 1s also furnished by the  McNally Pittsburgh
Manufacturing Corporation.  It 1s essentially  a rotating drum with a
cutter that takes a full stream cut once per revolution (Figure 4-9).
    All of these rotary samplers are usually Incorporated Into a sampling
system specified by the respective manufacturers.  A complete sampling
system generally Includes the primary sampler, secondary sampler,
crusher, and tertiary sampler.  This 1s known  as a three-stage sampling
system (Figure 4-10).  The primary Increment 1s a full cross-stream cut
from the total lot of coal.  The secondary Increment Is a full
cross-stream cut of the primary Increment.   After this stage, a crusher
1s used for sample reduction.  Sample division 1s accomplished by a
tertiary sampler which takes either a full  cross-stream cut of the
crushed secondary Increment, or splits the secondary Increment 1n half.
                                  39

-------
Figure 4-6.   Rotary plate divider.
                40

-------
                            GMD-500
                                            CONDENSED SPECIFICATIONS
                                             GMD-500 Self-Contained Rotary Sampler
                                             Drive:
                                             Speed:*
                                             Cutter Opening:
                                             Infeed:
                                             Capacity:
electric
18 in./sec.
0 = 2-1/4"
6" diameter pipe
25tph
to 3000 GPM
                                             GM-500 Self-Contalned Rotary Sampler
                                             Drive:
                                             Speed: *
                                             Cutter Opening:
                                             Infeed:
                                             Capacity:
                                             Particle Size:
 electric
 18 in./sec.
 0-1-1/4"
 3" diameter pipe
 500 Ibs. per hour
 •8 mesh
                                                  •NOTE: at center of cutter
                                                                    GM-500
Figure  4-7.   Ramsey sampler/Galigher  tertiary rotary.

                                41

-------
FEED
                      DENVER
                   VEZIN TYPE
                     SAMPLER
    REJECT   SAMPLE
                       2oH.P.,ll5 V., 60 Cy., SINGLE Ph.
                            GEARMOTQR-24 R.PM.
   FEED
                          10% CUTTER CAN
                           BE SUPPLIED
                         SAMPLE TRAY
                         CAP 80 Cu. In.;
                                            «i«
 PIPE,
 HOSE
ADAPTER
                           4-Z"DIA. MOUNTING HOLES

                           ON 9"DIA B C. 45° OFF
   Figure 4-8.  Denver \fezin sampler.

                  42

-------
Figure 4-9.  .  McNally rotary drum sampler.
                  43

-------
Figure 4-10.   Ramsey three stage sampling system
1 -  Primary sampler       6 -
2 -  Belt feeder           7 -
3 -  Bias test chute       8 -
4 -  Secondary sampler     9 -
5 -  Crusher
                    44
Sample chutes
Tertiary sampler
Sample collection
Reject

-------
Rotary plate samplers and rotary cutters are examples of tertiary
samplers.  Several sampling equipment manufacturers can supply a three-
stage sampling system.
4.1.4  Sample Reduction for Analysis
    For laboratory analysis, the sample must be further reduced to pass a
No. 60 (250 urn) sieve.  A pulverizer 1s used to grind the sample.
Suitable pulverizers Include the Braun laboratory pulverizer (Figure
4.11) and the Sturtevant sample grinder (F1g. 2.12), which can produce a
sample between 10 and 100 mesh.
    Suitable precautions should be taken to ensure a continuous steady
flow of coal Into the pulverizer to ensure uniformity of sample particle
size and composition.  The ground sample 1s then divided by use of a
small riffle (F1g. 4.13) until approximately 50 to 100 grams remain.
4.1.5  Sample Mixing
    To ensure the homogeneity of the 50 to 100 gram analytical sample, a
mechanical mixing device 1s employed.  A mixing wheel (F1g. 4.14) can be
used to rotate the sample containers so that the particles fall gently
from the top to the bottom of the container.  To obtain good mixing,
sample containers should be between one-half and two-thirds full and the
wheel should rotate at a slow speed.  To be most effective, samples
should be mixed Just before portions of the sample are extracted for
analysis.
4.2  SAMPLE ANALYSIS
    Proximate and ultimate analyses are often used to characterize
selected coal properties.  Proximate analysis refers to the determination
by prescribed methods of moisture, volatile matter, fixed carbon (by
difference), and ash.  Ultimate analysis Includes the determination of
carbon and hydrogen 1n the material (as found 1n the gaseous products
                                      45

-------
               •m.     \  - ^^tP~^ '^^
               •r*     V --^^^ '  *T.
               •**.     V-^~^  -=^~*
   Figure 4-11.   Braun lab pulverizer.
Figure 4-12.   Sturtevant  sanple grinder.
                   46

-------
                      .  J
Figure 4-13.   Shall lab riffle.
  Figure 4-14.   Mixing wheel.
             47

-------
of Us complete combustion),  sulfur,  nitrogen,  and  ash (1n  the materials
as a whole) and the estimation of oxygen by difference (ASTM 0121-78).
Although 1t 1s common to find several different methods for measuring
each parameter, 1t 1s generally through agreement between parties,  such
as the coal supplier and customer, that specific methods and operating
conditions are selected.
    The following description provides details  on the determination of
moisture, ash, total sulfur,  and gross calorific value (heat content) 1n
coal.
4.2.1  Determination of Moisture
    The determination of moisture 1n coal 1s calculated from the results
of two processes:  air drying and the determination of residual
moisture.  A1r drying 1s a partial drying of the sample to equilibrate  1t
with the room atmosphere 1n which sample reduction and division takes
place.. The total sample 1s weighed before being- put Into an air drying
oven or being spread onto a drying floor.  The  sample 1s dried until the
weight loss 1s not more than 0.1%/hour.  The percent air - dry loss 1s
calculated according to ASTM standard 03302 - 74^ ' superior type as:
                                  A = (L/6)100
where:
         A = air - dry loss,  %
         L = weight loss due to air drying
         G = weight of gross sample
A1r drying can be done as described earlier 1n  an air drying oven or on a
drying floor.
    The ASTM procedure for determining residual moisture requires the use
of a forced-air oven under strict temperature,  time, and air flow
                                      48

-------
specifications.   A one gram sample of  coal  1s  heated  1n  a  forced  air  oven
(F1g. 4-15)  at between 104°C and 110°C for  exactly  one hour.   The oven  must be
capable of renewing the heated air two to four times  each  minute.  The  air
current must be dried by passing through sulfuMc add or  another suitable
desslcant.  The percent residual moisture 1s  calculated  as follows:
                             M = [(A - B)/A]  x 100
where:
         H = % moisture
         A = Initial sample weight
         B - sample weight after heating
    Quality control measures to ensure the reliability of  moisture
content data begin with strict adherence to the standard method.   This
lessens the consideration of the method as  a  variable.  The reproduc-
1b1l1ty of results obtained by the same method, on  splits  of  the  same
sample; performed by different laboratories should  not differ by  more
than 0.50%, according to the ASTfT  .   A duplicate  sample check should
not differ by more than 0.30X.  Duplicate determinations should be
carried out frequently.  Other routine control checks should  Include
periodic verification of temperature settings and  preventive  maintenance
of equipment.
    Another method for moisture determination 1n coal 1s by the direct
gravimetric method.  In this procedure, the analysis  sample 1s heated In
a retort tube to 105°C-110°C for one hour.   A stream of pre-heated,
oxygen free, dry nitrogen 1s passed through the retort which  drives off
the moisture.  A weighing tube containing magnesium perchlorate,  or
another suitable desslcant, collects the moisture.   The weight Increase
of the desslcant tube 1s the moisture content of the sample.   The direct
gravimetric procedure for determining moisture content 1n coal 1s a
method recognized by the International Organization for Standardization
(ISO) as standard method ISO 331.(18)
                                     49

-------
Figure 4-15.   Forced air oven.
            50

-------
     ISO  requirements  for quality control 1n terms of precision are a
 difference  'of  less  than 0.20%  for  reprodudbHUy.  Exact procedures,
 Instrument  calibration checks, and preventive maintenance all contribute
 to Increased  reliability and should  be  Incorporated Into a quality
 control  plan.
                                                              / 1 Q\
     Another moisture  determination method  1s ISO standard 348V    , a direct
 volumetric  procedure  where  the coal  1s  distilled with  toluene, carrying the
 moisture to a  graduated receiver.  The  moisture condenses and separates from
 the toluene,  allowing direct calculation of the volume of water  released  from
 the coal.   An  advantage of  this method  1s  that the sample 1s  protected from
 oxidation.  A  sample  of 50  to  100  grams 1s needed.
     Quality control procedures for this method Include routine analysis
 of duplicates  which should  not differ by more than 0.5%.  A
 reproduc1b1!1ty  check 1s not an accurate Indicator of  reliability  since
 differing humidity  levels  1n different  labs would result 1n widely varied
 data.
     Although  the above methods are the  standards for moisture
 determination, there  are other accepted procedures.  Automatic coal
.analyzers are  being used for moisture determination.   The speed  of these
 systems  1s  a  great  advantage.  The Fisher  Model 490 Coal Analyzer
 performs moisture determinations  1n  an  oven with controls to  pre-set oven
 temperature and  drying time.   A microprocessor calculates percent
 moisture.   The LECO Corporation also has an automatically programmable
 furnace  suitable for  moisture  determinations.  The procedure  1s  similar
 to ASTH  methods. Duplicate determinations should also be run for  the
 automatic furnaces  techniques  to  determine repeating and reproducing
 reliability.
                                      51

-------
4.2.2  Ash Determination
    The determination of ash 1n a sample of  coal  1s  a  weight  loss
procedure.  The sample undergoes complete combustion,  and  the percent ash
1s calculated by weighing the residue.   The  ASTM  standard  D3174-73^6)
recommends two approaches using either  the dried  sample from  the moisture
determination or a new one-gram sample.
    When using the dried sample, a muffle Furnace (F1g. 4.16) 1s used to
gradually heat the sample to 500°C 1n one hour and 750°C 1n two hours.  The
sample 1s then heated to constant weight (t  0.001 gram) at 750°C.   The ash
content 1s calculated as a percent weight loss.
    When a new undMed sample 1s used.  1t should  be  gradually heated 1n a
cold muffle furnace until 1t glows red.   The heating rate  should be slow
enough to avoid the loss of volatile matter.  The combustion  should be
completed at 700°C to 750°C.  Before being weighed,  the sample must be
allowed to cool 1n a desiccator.
    Quality control measures to ensure the reliability of  ash
determination data are dependent on exact duplication  of the  procedures.
To ensure complete oxidation of the pyrltlc  sulfur and to  remove the
S0_, a large supply of air must be available 1n the muffle furnace.
The furnace should be capable of heating the sample slowly and of  having
Its temperature adjusted between 700°C and 750°C.  For samples with no
carbonates present, the ASTM repeatability limit  1s  0.2%.   It 1s 0.3% for
samples with carbonate.  Coals containing over 12% ash have a
repeatability limit of 0.5%.  The ASTM reprodudbllUy limits are  0.3%
for no carbonates present, 0.5% for coal with carbonates,  and 1.0% for
coals with more than 12% ash.
                (18)
    The ISO 117r    approach 1s similar but allows a faster  heating
rate and a higher maximum temperature.   A one- to two-gram sample  1s
heated to 500°C 1n 30 minutes and to 500°C to 815°C 1n 30 to 60 minutes.
The final temperature reached at 60 minutes  1s maintained  for an
additional 60 minutes.  The sample 1s then allowed to cool, weighed, and
                                 52

-------
Operating
tvmptriturt
2000 F/1093 C
2000 F/1093 C
2160F/1177C
2150 FM77 C
Chamber JIM,
W x H x 0"
4 x 3* x 9
4x3Xx9
4 x 3X x 9
4 x 3% x 9
Figure 4-16.   Muffle furnace.
             53

-------
returned to the 815°C muffle furnace for 15 minutes  or until  the sample
has attained a constant weight (+. 1.0 mg).   The ash  percentage 1s
determined from the weight loss.
    The precision levels set by the ISO as  a control are 0.2% for
repeatability and 0.3X for reproduc1b1Hty.  The same precautions
concerning procedural accuracy are necessary to ensure data reliability
for this method.
    More rapid methods are available for calculating ash content 1n coal
samples.  The Fisher Coal Analyzer and LECO Automatic Furnace use the
standard furnace method, but the heating rate, transition temperature,
final temperature, and duration of burn can be programmed Into a
microprocessor.  The results are displayed  as percent ash automatically.
Another automatic method 1s X-ray fluorescence:  The X-ray tube must be
calibrated with a known standard, and the sample 1s  compared against the
standard.  This method has not proved to be as accurate as the standard
                           (191
methods but 1s much quicker*   .
4.2.3  Determination of Total Sulfur
    There are three standard methods for determination of total sulfur 1n
the analysis sample of coal:  the Eschka method, the bomb washing method,
and the high-temperature combustion method^ '.
    The Eschka method requires 1.0 gram of  sample pulverized to pass No.
60 sieve.  The coal 1s mixed with 3.0 grams of Eschka mixture (2 parts
HgO, 1 part Na2(CO)_).  The mixture 1s placed 1n a porcelain or
platinum crucible and covered with one gram of Eschka mixture to prevent
sulfur lost as S0« during Ignition.  The crucible can then be heated
slowly over a low flame for 30 minutes (gradually Increasing the
temperature until all black particles are combusted) or heated 1n a cold
muffle furnace that reaches 800°C +. 25°C within one hour and maintained
at 800°C * 25°C until black particles are present.
                                    54

-------
    After combustion, the soluble sulfur compounds are leached out and
oxidized with bromine.  The sulfur 1s then precipitated as barium sulfate
(BaSO ).  The precipitate 1s filtered and then heated to 925eC until
the sample 1s at constant weight.  The ASTM requires a correction to  be
applied as part of this procedure because of the possibility of
solubility error for BaSO..  A standard sulfate should be analyzed 1n
the same manner as the analysis sample.  The correction factor 1s the
difference 1n weight of BaSO. precipitated from the standard and the
actual value for that standard.
                                                       (18)
    The Eschka method 1s also recognized by the ISO 334V   .  The
recommended procedure 1s the same as the ASTM method.  The ISO requires a
blank determination be conducted simultaneously with the analysis sample to
quantify any sulfur contamination from the atmosphere.  A known amount of
potassium sulfate 1s added to ensure a precipitate.  The amount of sulfur
detected 1n the blank, minus the known amount added, 1s considered as
atmospheric contamination.  This value 1s deducted from the result of the
analysis sample.  In addition, the ISO specifies the use of hydrogen  peroxide
to ensure complete oxidation of the sulfur to sulfate as opposed to the ASTM
requirement of bromine water.
    The percent sulfur content 1s calculated as:
                  (A - B) x 13.738
                         C
where:
         S = X sulfur
         A = weight of BaSO. precipitated (grams)
                           4
         B = weight of BaSO  correction (grams)
         C = weight of sample used (grams)
Besides the quality control methods of repeatability and reproduc1bH1ty,
a correction procedure must be run on a standard sulfate as part of the
                                   55

-------
control.   The correction procedure  must  be  an  exact  duplication  of  the  sulfur
determination procedure.  A reagent blank determination  can  be  run  to quantify
the amount of sulfur contamination  from  the atmosphere.  .The precision  limits
for repeatability are 0.0554 for coals  containing less  than  2% sulfur, and
0.10X for higher sulfur coals.   The reproduc1bH1ty  limits  are  0.10% for  less
than 2X sulfur coal  and 0.20% for higher sulfur coals.
    The ASTM's bomb washing method  uses  the oxygen-bomb  calorimeter
washings from the determination of  calorific value to  calculate total
sulfur content'  .   The Interior of the  oxygen bomb  (F1g.  4.17)  1s
washed with a spray of water containing  methyl orange  until  no  add
reaction 1s observed.  The washings are  collected, and an  add  correction
tltratlon 1s performed for calorific value  determination.   To precipitate
the sulfur, NH.OH 1s used to reduce the  pH  to 5.5 to 7.0.   The  washings
              4
are then boiled and filtered.  This residue 1s washed  with  hot  water
several times, and then 1 ml of bromine  water 1s added and  enough HC1 to
make the solution slightly acidic.   The  remaining steps  of  adjusting the
addlty, precipitating the BaSO., and determining total  sulfur  content
are the same as 1n the Eschka method.
    The precision limits for repeatability  and reproduc1b1!1ty  are  the
same as the limits given by the ASTM for the Eschka  method.   Additional
quality control techniques should be used such as a  reagent blank
determination and a standard sulfate determination.
    Another sulfur determination recommended by the  ASTM 1s the
high-temperature combustion method    .  A 0.5 gram sample  1s burned 1n
a tube furnace (F1g. 4.18) under a stream of oxygen  to prevent formation
of nitrogen oxides.  The oven temperature 1s 1350°C, with  the sample
moving progressively closer to the hot zone to ensure a slow heating
rate.  After ten minutes, the sample 1s  1n  the center of the hot zone and
remains there for four minutes.  Sulfur  and chlorine oxides produced are
collected 1n two absorption bottles containing hydrogen peroxide producing
                                      56

-------
Figure 4-17..  Adiabatic bomb calorimeter.
       Figure 4-18.   TUbe furnace.
                57

-------
hydrochloric and sulfurlc adds.  The total sulfur 1s calculated from the
total acidity which 1s determined by tltratlon with standard NaOH and
then with standard H0S(L.
                    i  4
    The same procedures and equipment are described 1n the ISO standard
351 for determination of total sulfur by the high temperature combustion
      / -I g\
method    .   A furnace temperature of 1250°C and 1350°C can be used.
The higher temperature allows a lower oxygen flow rate without the
formation of nitrogen oxides.  At the higher temperature,  the sample
should be covered with 0.5 grams of aluminum oxide.  At 1250°C, 0.5 grams
of kaolin or 0.15 grams of Iron phosphate are used.
    Strict adherence to the standard procedures, Including standard
reagents, 1s Important for the control of this method.  Results of
duplicate determinations performed by the same lab should  not differ  by
more than 0.05X.  Reproduc1b1l1ty experiments performed or duplicates by
different laboratories should not differ by more than 0.15X for coals
containing less than 2X sulfur, and by more than 0.25X for higher sulfur
coals.
    There are many methods, besides the accepted standard  methods, used
by Industry.  Automatic analyses are widely preferred because of the
short analysis time, but they are not well recognized for  the reliability
of their data.
    A semi-automated procedure which the LECO Corporation  Introduced  many
years ago used a high temperature Induction furnace to combust the coal
and convert the sulfur to SO  1n an oxygen rich atmosphere.  The
evolved SO  was automatically directed to a glass chamber  where 1t was
1odometr1cally titrated with standard sodium thlosulfate.   This method
has been replaced by a new LECO technique which uses a microprocessor to
Initially weigh the sample, control the combustion process, and calculate
final sulfur concentration.  This technique 1s covered by  ASTM D-4239-83,
"Sulfur 1n the Analysis Sample of Coal and Coke Using High Temperature
Tube Furnace Combustion Hethod."  This method 1s similar to the ASTM
                                    58

-------
0-3177-75 Method C high temperature combustion procedure.   In the new
high temperature tube furnace method, the sample 1s placed 1n a ceramic
boat, weighed, and Inserted Into a resistance wire heated  furnace.   The
evolved SCL 1s measured spectroscoplcally by an Infrared detector and
compared to calibration standards automatically.  Total  analysis take two
minutes.
    Another automatic analyzer based on the high temperature combustion
method 1s the Prelser/MINECO sulfur analyzer.  Although  not as fast as
the spectroscoplc method, 1t offers excellent precision  when compared
with the Eschka method.
    Another accurate and rapid system 1s the Fisher Sulfur Analyzer
System.  The sample 1s burned 1n a combustion-tube furnace.  The oxidized
sulfur compounds are absorbed by a diluent, resulting 1n a decrease 1n
current 1n a preset platinum electrode.  THrant 1s automatically added
to restore the current to the preset level.  The amount  of tHrant needed
Is directly related to sulfur content.  Analysis time 1s approximately
two minutes.
    Another combustion tube type of analyzer has a different sulfur detection
scheme.  The Perkln-Elmer Corporation's thermal-conductivity analyzer
determines sulfur concentration by reading the differential response across
two thermal conductivity cells.  Accuracy 1s reported to be 0.3X of total
sulfur.
    Several companies make X-ray fluorescence analyzers  capable of
measuring total sulfur content 1n coal.  The operating principle 1s
generally the same.  Low level radiation 1s emitted from an Iron
excitation source, causing X-rays that are characteristic  of sulfur.
These X-rays are measured and are proportional to the sulfur
concentration.
    To ensure reliability of the data obtained from rapid  sulfur
analyzing systems, the manufacturer's recommendations for  equipment
                                    59

-------
maintenance, standardization and calibration,  and. operating procedures
should be closely followed.  In addition,  results' of duplicate analyses
should be within accepted repeatability and reproduc1b1Hty limits for
the determination of sulfur.
4.2.4  Determination of Calorific Value
    The calorific value of coal 1s determined  1n a bomb calorimeter.  The
static Isothermal and adlabatlc methods are two accepted standard
procedures.  In the static Isothermal method,  a water Jacket 1s used to
protect the calorimeter from the effects of external heat.  In the
adlabatlc method,.the jacket temperature 1s continually adjusted
throughout the procedure to match the temperature of the calorimeter,
thereby minimizing heat exchange between the bomb and the surroundings.
Because of this lack of heat transfer, no correction factor 1s needed 1n
the adlabatlc method.  In both cases the gross calorific value 1s
determined.  The gross calorific value 1s the amount of heat produced by
complete combustion of a specified quantity of coal under strictly
specified conditions 1n an' oxygen bomb calorimeter.  The net calorific
value.1s a lower value and can be calculated from gross calorific value.
A suitable oxygen bomb calorimeter 1s shown as F1g. 4.17.
    The procedure for the static Isothermal method of determining gross
calorific value Includes a standardization procedure for calibrating the
calorimeter.  Ten separate calorific value determinations are run using
benzole add pellets as the standard fuel source.  The pellet weight
should result 1n the same temperature rise as  1s expected with the coal
to be tested.  This allows the calorimeter's energy equivalent to be
calculated.  The bomb should be charged to the same oxygen pressure for
each calibration and each calorific valve determination.
    The calorific value 1s determined by burning a one-gram sample and
observing the temperature rise of the calorimeter.  The corrected
temperature rise 1s calculated by making radiation corrections for heat
                                      60

-------
contributed by other sources.   The gross calorific value (Btu/lb)  1s then
obtained by multiplying the corrected temperature rise by the energy
equivalent which 1s then adjusted to account for energy formation  from
side reactions.  The ASTM and  ISO recognize the static Isothermal  method
1n their standards ASTM D 3286-77 and ISO 1928{6>18).
    The adlabatlc bomb calorimeter method 1s a similar process except
that 1t requires control of the water Jacket temperature.  A
standardization procedure 1s used to obtain the water  equivalent of the
calorimeter using benzole add as the standard fuel  source.  The pellet
weight should result 1n the same temperature rise as would be expected
for the analysis sample.  The  procedure should be exactly the same as for
the calorific value determination.
    A T.0-gram sample (i 0.1 mg) 1s placed 1n the crucible, and 1.0 ml of
water 1s added to the bomb.  A measured length of firing wire 1s
connected to the Ignition terminals.  The bomb 1s charged to an oxygen
pressure between 20 and 30 atm.  The pressure must be  consistent for each
calorific determination.  During combustion, the Jacket temperature must
be continually adjusted to equal that of the oxygen bomb.  The gross
calorific value 1s calculated  as the corrected temperature rise
multiplied by the water equivalent, minus corrections  for heat added by
other sources, divided by the sample weight.  ASTM standard 02015-77^ '
                     (18)
and ISO standard 1928     are accepted procedures for  determining gross
calorific value by the adlabatlc bomb calorimeter.
    Quality control measures to ensure the reliability of data from the
calorific value determinations Include frequent calibration of the
calorimeter, maintenance checks of equipment, and exact procedures.  The
repeatability checks should not differ by more than 50 Btu/lb.  The
reproduc1bH1ty checks should  not differ by more than  100 Btu/lb.
                                   61

-------
    Other methods besides the bomb calorimeters are available for
measuring calorific value.  Several X-ray fluorescence systems have been
developed which measure a radiation level which 1s proportional to Btu.
The rapid analytical results available from these systems make them more
advantageous, but calorific value determination by bomb calorimeter 1s a
more proven method.
    Automatic, microprocessor controlled, bomb calorimeters are available
from the LECO Corporation and the Fisher Scientific Company.  The entire
process 1s preprogrammed and totally automatic.  Both procedures are
based on ASTM D 2015, the adlabatlc bomb calorimeter method.
    Manufacturers' specifications and Instructions should be closely
followed for all automatic systems.  Calibration checks should be
performed on a regular basis as should duplicate sample runs.  The
precision values resulting from repeatability and reprodudblHty tests
should be within specified limits for the adlabatlc bomb calorimeter
method'.
    The heating value of coal may be computed using empirical
    itlonshlps and the concentratl
Dulong-Petlt formula 1s equal to:
relationships and the concentrations  of various  constituents^    .   The
                  Btu/lb = 145.44 C + 620.28 (H-0/8) + 40.50 S

Where:  C = carbon (% by weight - as received)
        H = hydrogen (% by weight - as received)
        0 = oxygen (% by weight - as received)
        S = sulfur (% by weight - as received)
     Other relationships have been developed which may Improve the accuracy
of predicting the as-received heating values for certain types of coal.
These Include the Grummel-Davles formula and the T.K. Subramanlan formula.
                                      62

-------
                                SECTION 5
                SULFUR LOSS IN COAL-FIRED UTILITY BOILERS

     Coal feedstock 1s usually sampled as 1t 1s bunkered.  When 1t 1s
removed, 1t moves on to the pulverizers and thence to the boilers.
Sulfur 1s removed during pulverizing and combustion from pathways other
than the emission of gaseous sulfur dioxide 1n the flue gas.  Hence the
amount of gaseous sulfur dioxide 1n the flue gas 1s not simply related to
the measured sulfur 1n a CSA program.  As sulfur dioxide 1s regulated by
EPA, guidelines need to be developed to estimate the sulfur dioxide based
on as-bunkered coal sulfur measurements.  Several factors must be
considered 1n this development:   the coal characteristics (ash
constituents, organic and pyrltlc sulfur content, and heating value), the
combustion process (boiler size, firing type, and generating load), and
possible fuel additives.
     Information on sulfur loss  has been collected by different utilities
for specific boiler units; however most of.this Information has not been
published.  Other reports that address this subject on a more
comprehensive basis are limited  1n scope.  Therefore, the verbal  comments
of some of the utility sources contacted for this project have been used
as the basis for this section.
     Mass balance calculations for sulfur are extremely difficult to
perform.  One difficulty 1s that some of the sulfur 1s released as
volatile compounds (which are Inherently difficult to sample).  A second
1s the difficulty 1n achieving truly representative samples of the coal
entering the bunker, the pulverized coal entering the boiler, bottom ash
leaving the boiler, fly ash collected by the emission control devices,
and fly ash emitted to the atmosphere.  A third 1s the difficulty
associating a fixed coal lot sampled as-bunkered to the as-fired  samples
                                    63

-------
since the time between these samples 1s a random variable depending on
the plant's bunker retention volume as well as the coal handling and
storage system.  A final problem 1s the difficulty 1n accurately
measuring the mass of material entering and leaving the boiler
(combustion air, coal, bottom ash, fly ash, and flue gas).
5.1  SULFUR LOSS IN PULVERIZERS
     Coal contains sulfur 1n three forms:  pyrltlc sulfur, contained 1n
discrete particles or bands of sulflde minerals; organic sulfur, bound 1n
the chemical structure of the coal; and sulfate sulfur, an oxidation
product^   .  Sulfate sulfur, usually as calcium or Iron sulfate, 1s
                                                (22)
present only 1n concentrations below 0.1 percent   '.  Organic and
pyrltlc sulfur are both capable of being oxydlzed  to gaseous sulfur
oxides during combustion.  The organic sulfur cannot be removed by
mechanical means whereas some of the pyrltlc sulfur can be removed by the
pulverizers.
     The pulverizers crush the coal prior to firing.  They have the
capability of rejecting particles that are denser  than coal.  This
material, known as mill rejects, typically consists of rock fragments,
larger chunks of pyrlte, and'tramp metal.  Eastern bituminous coals have
high pyrltlc sulfur content 1n contrast to sub-bituminous and I1gn1t1c
                   (22)
coals from the west    .  Thus, eastern coals contain sulfur 1n a form
                                                             (23)
able to be rejected by the pulverizers.  A mass-balance study     of a
plant using western Kentucky coal showed a reduction of 0.3 percent;
several utility representatives thought this figure typical, with maximal
reduction 1 percent.
5.2  SULFUR LOSS IN COMBUSTION
     During coal combustion, a small amount (typically 0.5 to
                     •
1.5 percent) of fuel sulfur 1s converted to SO-, although the high
flame temperatures and long residence times of modern utility boilers
                           (24)
favor the formation of S02>      The proportion of SO- to S03 1s
                                    64

-------
controlled by the combustion area temperature,  the amount of excess air
                                                            (21)
supplied to the flame, and the presence of certain catalysts    .

     The most significant sulfur loss during coal  combustion comes from

reaction with coal-ash constituents and/or fuel  additives (e.g.,

limestone) to form sulfates.  Estimates range from 5 to 40 percent of the
                          (25)
total sulfur would be lostx  ' 1n the process.

5.3  GUIDELINES

     In general,  not enough Investigation has been done 1n considering

the effects and Interrelationships of the various  factors Involved 1n

sulfur retention  to allow a blanket method to be suggested.   However,

some possibilities can be given.

     a.   A constant 5 percent rate seems a reasonable compromise  between
          regulatory demands for conservancy and Industrial  demands for
          "reality".

     b.   Each utility can take a constant rate  Csay 5 percent)  or can,
          using mass balance,  prove 1t 1s entitled to more.   The
          regulatory difficulty here 1s that the retention rate  1s so
          variable that past Information may bear  little relation  to the
       .   future.

     c.   Allow no retention rate.  This 1s the  most conservative
          approach forcing the sources to bear  the entire burden of the
          retention 1f they wish to use CSA.

     d.   Allow no retention 1n general but allow  each source,  using mass
          balance, to prove 1t 1s entitled to one.  The drawback  stated
          1n (b)  applies.
                                    65

-------
                                 SECTION 6

                 COMMON COAL INDUSTRY AND UTILITY PRACTICES
                       FOR COAL SAMPLING AND ANALYSIS
     Several common coal sampling and analytical techniques are currently
 used by coal-fired utilities to evaluate their fuel.  The purpose of this
 section 1s to determine the extent of use of these methods and any
 quality-control measures associated with them.  This Information 1s
 Important to the study because 1t Indicates how closely current Industry
 practices match the proposed EPA Reference Method 19A (48 PR 48960, Oct.
 21, 1983) requirements and because 1t may show where alternative
 procedures need to be Incorporated or approved for use somewhere within
                                           (26)
 the proposed regulation.  A previous study     estimated that only 16%
 of the 90 utilities surveyed had existing sampling and analytical
 procedures that complied with the requirements of the proposed
f regulation.  This section does not attempt to compare existing practices
 with those requirements but only Identifies what the more commonly used
 practices are.
 6.1  INFORMATION SOURCES
     One of two major Information sources was a report entitled "Electric
 Utility Coal Sampling and Analysis Practices:  A Comparison to Proposed
 EPA Reference Method 19A Requirements Based on Utility Responses to FERC
        / pt \
 Survey"    .  This report tabulated and analyzed certain data collected
 from a survey of 190 utility plants for coal sampling and analysis
 Information.  The tabulations were set up so they could be compared with
 the proposed Reference Method requirements; therefore, the sampling
 method was not specifically defined beyond a group of ASTM methods.  The
 sampling location was not specified other than "as-received" or
 "as-f1red," and the method of sulfur analysis was not defined 1f any
 method other than the ASTM standards was used.
                                    -66-

-------
    The other source of Information was Individual  contacts at 24 utility
plants that Versar had previously worked with.   These plants were part of
a nationwide sampling program and, as such,  represent a variety of boiler
types, coal types, and geographic locations.
    Since Information was solicited from Individual contacts, more
specific Information was obtained than was presented 1n the Baker
      / OR)
reporv   .  Also, a few Inconsistencies between the FERC survey data
and the Individual contacts were found for some plants.  These
Inconsistencies (e.g., type of sulfur analysis  method used) were resolved
1n favor of the responses of the Individual  contacts.
    Some of the 24 plants contacted by Versar were also Included 1n the
FERC survey.  An effort has been made to display these two sources
separately.  Therefore, the duplicates were deleted from the FERC data
base, thus decreasing the total number of plants represented by that
source.
6.2  SAMPLING TECHNIQUES
    There are several coal sampling techniques  1n use at different
utilities.  The choice of technique generally 1s based on either the
                                                                      •
purpose of the coal sampling program (e.g.,  evaluating vendor
performance; monitoring coal sulfur Input to boiler) or the design of the
plant and adaptability of sampling equipment.
    The techniques 1n current use are described 1n ASTM standard method
02234-76 (Collection of A Gross Sample of Coal) and are divided Into Type
I and Type II methods(6).
6.2.1  Type I
    Type I methods are those not subject to discretionary bias 1n
selecting sample Increments.  These methods Implicitly Involve some sort
of automated equipment and have the code designation I.  Sample
Increments can be removed from a stopped belt (Code A), the entire width
                                      67

-------
of a moving belt (Code B), a partial  width of a moving belt (Code C), or
a stationary source (Code 0).  The sample Increments can be evenly spaced
(Code 1) or random (Code 2).  Thus, a sample of coal composed of
evenly-spaced Increments taken from the entire width of a moving belt by
an automatic sampling device would be designated IB1.   Type I, Method B
1s typically used where coal 1s transported around a plant by belts and
1s the method of choice when as-fired coal quality 1s  to be monitored.
The sampling equipment can be Installed at the end of  a belt (e.g, where
the belt changes direction).  Method IA 1s less frequently used because
1t Involves stopping the belt.  Interruptions of this  sort are not
usually tolerated at electric utilities.  Method 1C 1s also less common
because most automatic belt-sampling equipment Incorporates a full-stream
cut application 1n Its design.  Method ID1 or 2 1s not commonly used.
6.2.2  Type II
    Type II methods Involve some measure of human discretion 1n selecting
sample Increments and can be used on any of the sources (Codes A-D)
listed above.  Increments can be evenly or randomly spaced (Codes 1 and
2) as before.  Thus, manually-sampled Increments taken at random times
from a coal pile would be designated 1102.  Type II methods are used 1n a •
wider variety of situations.  For example, taking samples from the body
of a truck or rail care delivering the coal and manually scooping samples
from a belt Inside the plant are Type II.  Type II, Method D 1s more
frequently used when coal vendor performance 1s being checked; Methods B
and C are more common when evaluating as-fired coal quality by manual
methods.
6.2.3  Survey Results
    The sampling methods employed at several utility plants are shown 1n
Figure 6-1.  These data represent a combination of results from the
Baker^  ' report 1n which sampling methods were characterized 1n one of
two ways (Methods IA1-IB1-IC1 or "other") they Include responses from 24
                                        68

-------
            ASTM 02234
           METHOD CODE !
                                                                                                    % OF TOTAL
UD
        IA1-2

        IB1-2

        IC1-2

        101-2
  undifferentiated
  IA1.IB1.orld

       IIA1-2

       IIB1-2

       IIC1-2

       IID1-2

method other than
IA1.IB1. orlCI
                        I
                                                           (9)
                                  (2)
                                                                                                    (120)
                 II DATA FROM
           I  II  II VERSAR CONTACTS
                                                                                                    DATA FROM
                                                                                                    BAKER, 1982
                                                                                                                 6.2
                                                                                                                 61.5
                               4.6

                               1.0

                               26.8
                                                            10
                                                          12
SO
60
100
110
120
130
                         NUMBER OF PLANTS (TOTAL = 195)
                      — DOES NOT INCLUDE ONE PLANT SAMPLING COAL WITH CYCLONES IN PULVERIZER
                         OUTLET PIPES (MODIFICATION OF ASTM0197).
                                           Figure 6-1.FREQUENCY OF SAMPLING METHOD

-------
utilities that Versar had previously been 1n contact with.   The raw data
do not differentiate between even or random Increment spacing.   However,
based on the nature of the sampling equipment and procedure 1n  each case,
1t 1s likely that all samples taken with automatic equipment (Type I) are
evenly spaced and that those taken manually (Type II) are randomly
spaced, unless the utility 1s extremely rigorous 1n Its application of
standard methods.
    Versar's experience with power plant operations Indicates that
stopped-belt, partial-stream cut, and stationary sampling methods are
almost never used when automatic sampling equipment 1s 1n place.  This 1s
reflected 1n the data collected from Versar contacts, and 1t Indicates
that a large percentage of the "undlfferentlated" Type I responses shown
1n Figure 6-1 probably refers to Method IB1.  Similarly, the Versar data
Indicate that 1f Type II Methods are used, sources C and 0 are the most
common with the emphasis on belt sampling.  It can be assumed that a
fairly-large percentage of "other" sampling methods consist of these
sources as well.
6.3  SAMPLING LOCATION
    Coal can be sampled at many locations  In a power plant depending on*
the method of coal handling, accessibility, and sampling objective.  Most
frequently, coal 1s  sampled from a conveyor belt because this affords
good accessibility and brings the coal to  the sampling equipment rather
than vice versa.  Some operators, however, take samples from delivery
vehicles to monitor  Incoming coal quality  or from bunker storage 1n  the
plant.  The most Important difference between locations 1s whether the
sample represents "as-received" or "as-f1red" coal.
    As-fired samples are taken at any location after the storage bunker
outlet.  These  samples are frequently taken from the coal feeders or from
a clean-out pipe on  the bunker line going  to the feeders.  Samples are
usually not taken from the pressurized lines after being pulverized,
although one plant contacted did this with the aid of  1n-l1ne cyclone
samplers.

                                      70

-------
    As-received samples are collected from any point before the bunker
outlet.  The main difference between as-received and as-fired samples 1s
that the as-received samples represent coal that will be fired at some
time after being retained 1n the bunkers or storage piles and after
passing through the coal-handling system.  The retention time caused by
this storage as well as any blending which occurs depends on coal
handling system characteristics and boiler operating parameters; 1t can
vary significantly.
    The relative frequency of sampling locations from 195 utility plants
1s shown 1n Figure 6-2.

6.4  BIAS TESTING/SAMPLING QUALITY CONTROL
    Bias testing of sampling methods Involves determining the difference
(usually on a statistical basis) between samples taken by the method or
equipment to be tested and those taken by some reference method.  The
most widely-used reference method for these comparisons 1s the method for
collection of a gross sample of coal (ASTH 02234-76, Appendix 1)'  .
Other tests have been formulated by manufacturers of sampling equipment,
but these are not prevalent (see Section 3.0).
    Discussions with several utility plant representatives about sampling
methods used at each plant Indicated that bias testing 1s not routinely
performed, and may be done only 1f the plant 1s undergoing some sort of
performance or compliance test, 1f at all.  The utility representatives
also Indicated that routine quality control was not performed at most
plants other than to ensure that automated sampling equipment was
properly maintained and adjusted or that standard manual sampling
procedures were followed during dally operations.

6.5  ANALYTICAL PROCEDURES
6.5.1   Sulfur Analysis
    Several analytical methods are available for the analysis of coal
characteristics, particularly for sulfur.  The sulfur analytical methods

-------
           LOCATION
BEFORE BUNKER
(ASREC'D)
FROM BELT
FROM STATIONARY
SOURCE

.OTHER LOCATION

AFTER BUNKER (AS FIRED)

UNSPECIFIED
                                         85
                                          94
                                    02468

                                    NUMBER OF PLANTS (TOTAL = 196)
                                                      10
50
                                                                                             % OF TOTAL
70
                    80
90
                                    43.4

                                    1.0

                                    3.6

                                    4.1



                                    48.0
100
                                        Figure 6-2.  SUMMARY OF SAMPLING LOCATIONS

-------
can be roughly divided Into standard wet chemical  techniques and
automatic photometric techniques.  The wet chemical  techniques are
generally the standard reference methods, and they are described 1n ASTM
D3177-75, Test Method for Total Sulfur In the Analysis Sample of Coal and
Coke   .   The photometric techniques are basically modifications of
these standard techniques and employ colorlmetrlc, turb1d1metr1c, or
Infrared  absorption to quantify the amount of sulfur evolved by the
various analytical preparation procedures (usually combustion).
    The results of the survey of sulfur determination techniques are
presented 1n Figure 6-3.  The data Indicate an almost even split between
standard  wet chemical methods and automated analysis techniques.  Note
that the  data from the Baker (1982)(26) report listed as "other" were
assumed to refer to automated photometric techniques since the other
categories listed were very specific.
6.5.2  Heat Content. Moisture. Ash Analysis
    The heat content of coal samples 1s determined with either an
adlabatlc calorimeter ASTM Method 02015-77, or an Isothermal calorimeter
ASTM Method D3286(6). Nearly all the utility plants surveyed (184 out
of 196, or 93.9%) determined the heat content of their coal samples with
the Parr  adlabatlc colorimeter, using Method 02015.  Four plants (2.0%)
reported  using the other method.
    Residual moisture determinations were listed two ways 1n the Baker
(1982) report.  Those conducted according to ASTM Method 03173 or those
conducted by grinding the coal to 60-mesh size but performing some test
other than that specified 1n D3173.  The results show that 85.3% of the
plants surveyed reported using 03173.  Another method was reported to be
used by 4.1%; 20 utility plants did not 11st any method.
    Total moisture Information was derived by using ASTM 03302 on 60-mesh
coal (67.4% of respondents), by a test other than 03302 on 60-mesh coal
(6.8% of  respondents) or by an entirely different method (23.3% of
respondents); five utility plants did not report.

                                  73

-------
ANALYSIS METHOD
                        54 OF TOTAL
ASTM
D3177
"OTHER
INFRAR
COLORI
COMBIN
TWO ME
(NOT RE
' ESCHKA
BOMB WASH
COMBUSTION
"*
ED DET.
METRIC
ATION OF
THODS
PORTED)
(23) |
^1
«> (237) |


^s^»
^-
1 (5)
1(6)

(54) I

(71) |
| | DATA FROM BAKER (1982)
[|\\\| VERSARDATA
11.7
27.6
11.7
36.2
4.6
2.6
2.6
3.1
                 0      S     10     IS     20
                 NUMBER OF PLANTS (TOTAL = 196)
25
40
50
60
70
      80
                      •"OTHER" CATEGORY IN BAKER (1982) MOST PROBABLY REFERS TO
                      A COLORIMETRIC OR PHOTOMETRIC AUTOMATED ANALYSIS METHOD
                   Figure  6-3. FREQUENCY OF SULFUR ANALYSIS METHOD

-------
    Information about ash analysis was not collected from any of the
Information sources; however, 1t seems likely that since the majority of
stations used ASTM standard methods for moisture and heat content
analyses, a similar majority would use the appropriate standard method
for the analysis of ash.
6.6  ANALYTICAL QUALITY CONTROL
    Most of the utilities contacted Indicated that they practice fairly
extensive quality control procedures,  ranging from analyzing standard
samples of coal with each batch of regular analyses to participating 1n
"round-robin" programs with other laboratories.   The most frequent
response was that standard samples are run on a  dally, every-other-day,
or weekly basis, often 1n combination  with batch standards, blanks, or
duplicates.  Only a few utilities reported conducting QC procedures with
a frequency of more than one month.
6.7  SUMMARY
    The above discussion Indicates that the following sampling and
analytical procedures are common 1n current utility practice:
    1. Sampling from conveyor belts using automated full-stream cut
       equipment 1s used by approximately 50% of all the plants
       considered by Versar, and probably 67% of all the plants
       considered 1n this survey.
    2. Sampling from a conveyor by taking random manual grab samples 1s
       used by 38% of the plants contacted by Versar, and probably 36% of
       all the plants considered.
    3. Sampling coal to define as-received quality 1s used by 43% of all
       plants considered as opposed to 4% taking as-fired samples;
       however, no location was specified for 48% of the plants.
    4. Determining heat content with an adlabatlc calorimeter 1s used by
       74% of all plants considered.
    5. Analyzing for sulfur by the bomb-washing  method 1s by used  54% of
       those plants reporting that wet chemical  procedures are used.
                                   -75-

-------
6. Analyzing for sulfur using automated,  Infrared-detector-equipped
   analyzers may be more prevalent than use of standard wet-chemical
   procedures:

7. Using ASTM standard methods for the analysis of residual  moisture,
   total moisture,  and ash 1s universal.

8. Calibrating equipment with standard samples, duplicates,  or
   blanks, 1s done most frequently on a dally, every-other-day,  or
   weekly basis.

9. Analyzing standard samples 1s done for each batch (usually 5  to 10
   samples).
                               -76-

-------
                                SECTION 7
                 THE IMPORTANCE OF COAL SULFUR PARAMETERS

7.1  TIME SERIES
     The problem of predicting what will happen 1n the future 1s
extremely Important because many of the decisions we make 1n our everyday
lives are based on predictions of conditions that are likely to exist 1n
the future.  Predictions of future events are called forecasts and the
act of making such predictions 1s called forecasting.
     Clearly, forecasting 1s very Important 1n many types of
organizations since the predictions of the future must be Incorporated
Into the decision making process.  For example the government of a
country must be able to forecast such things as air quality, water
quality, unemployment rate, Inflation rate, and welfare payments 1n order
to formulate Its policies.
     Since accurate forecasts are required 1n so many decision making
forecasts, a forecaster must rely on Information concerning events that
have occurred 1n the past; that 1s, 1n order to prepare a forecast, the
forecaster must analyze past data 1n order to Identify a pattern that can
be used to describe 1t.  Then this pattern 1s extrapolated, or extended.
Into the future 1n order to prepare a forecast.
     Time series data are the most widely used data for forecasting.  A
time series 1s a sequence of measurements of the same variable made at
different times.  Equally spaced time points are used 1n most time series
studies, although they are not always used.  Business time series often
Involve yearly, quarterly, or monthly observations, but any other time
period may be used.
     There are many examples of time series data, for example:  sales of
a particular product over time, dally mean temperature, over time, and air
                                   -77-

-------
or water quality over time.  In particular, dally SO  emissions data
are a time series.
7.2  COAL TIME SERIES
     The measured values of dally sulfur content, ash content, and
heating value for coal are examples of time series.  We will denote by C
the "true" value for the coal property being measured.  Unfortunately, we
do not exactly report C since the measuring process 1s not exact.  If we
let X be the reported data, then symbolically we may write
                                X = C + HE
where ME 1s the possible error Introduced by the measuring process.
This latter term consists of variations 1n the sampling and analysis
procedures.  It Incorporates possible errors 1n representing.the entire
coal or gas stream by a single laboratory sample and 1n the procedures
used to obtain that single sample.  It also Incorporates possible human
errors In the analytical procedures or 1n reading output from automated
analytic equipment.  Last, 1t also Incorporates possible equipment
malfunction.  We assume that the level of ME 1s unrelated to that of the
actual process C and that, on average, ME 1s zero.  Of course, for any
particular measurement, ME probably will not be zero.
7.3  VARIABILITY OF COAL PARAMETERS
     The variance 1s the average of the squares of the deviations of the
observations from their mean.  Therefore, the variance makes sense as a
measure of spread or variability around the mean.  To understand this,
note that 1f the observations are spread out, they will tend to be far
from their mean, both above and below.  Some deviations will be large
positive numbers, and some will be large negative numbers.  But the
squared deviations will all be large and positive, so the variance will
be large when the data are spread out and small when the data are close
together.
                                   -78-

-------
7.3.1  General Characteristics
     Variability 1n coal Is an Inherent quality which 1s a function of
the geologic and chemical processes of Us formation.  Coal originated
from the deposition of plant materials 1n swampy areas at various time
Intervals during the last 50 to 350 million years.  These depositions
have resulted 1n more or less horizontal strata which we denote as
seams.  Subsequent geological activities have distorted these layers to
varying degrees.  Thicknesses vary from several Inches to several hundred
feet.
     Although coals within a given seam are more consistent 1n their
properties than coals from different seams, substantial variations 1n
coal properties exist within larger seams both vertically and
horizontally.  The statistical treatment of coal data must recognize both
the spatial relations (correlation) as well as the random elements of the
samples.
7.3.2  Importance of Variability
     The ability of a coal to pass emission standards depends on the mean
level as well as the variance and correlation (see 7.4 below) of Us
sulfur content.  To see the effect of the variance, let us consider the
hypothetical case of a source burning 1.0 pounds of SO. per MMBtu coal
with a 1.2 pounds of SO  per MMBtu standard.   If the variance were
zero, then each data point would be 1.0 and the source would clearly be
1n compliance always.  However,  there 1s variation 1n true data.  If the
data were normally distributed,  then about 16 percent of the data lies
above one, 2 1/2 percent above two and 0.13 percent above three standard
deviations (the standard deviation 1s the square root of the variance) of
the mean.  So suppose we have a  fairly typical coal:  1t would have, say,
a standard deviation of 0.1.  Then data above two standard deviations of
the mean 1 * 2(.l) =1.2 would not be complying with the standard (see
Figure 7-1).  This would occur approximately  2 1/2 percent of the time.
If the coal were more uniform (say from a dedicated mine), It might have
a standard deviation of, say, 0.07.  This uniform coal would be 1n

                                   -79-

-------
STANDARD DEVIATION 0.1
                                           STANDARD 1.2
                                          MEAN 1.0
STANDARD DEVIATION 0.07
                                           STANDARD 1.2
                                      *=r—  MEAN 1.0
STANDARD DEVIATION 0.2
                  r\
       \
                                           STANDARD 1.2
                                           MEAN 1.0
        FIGURE. 7-1.  EFFECT OF VARIANCE ON S02 EMISSIONS






                              80

-------
non-compllance about 0.1  percent of the time since 1.2 1s about three
standard deviations above the mean.  Finally,  1f the coal were purchased
on the spot market, 1t might easily have a standard deviation of 0.2.  In
that case,  1t would be 1n non-compllance about 16 percent of the time
(1.2 1s only one standard deviation above the  mean).
7.3.3  Reducing Coal Sulfur Variability
     Reduced sulfur variability 1s desired for an enhanced probability of
complying with emissions  standards.  Processes such as coal cleaning,
blending, and mixing all  contribute to reducing variability.
     Physical coal cleaning 1s a process where the coal 1s crushed so
mineral Impurities can be released.  The Impurities are washed out by
specific gravity separation.  Although originally Intended only to
produce a lower ash coal, the process has proved to be a good sulfur
reduction step (removing a reasonable percentage of the pyrltlc sulfur)
and a good method of producing a more homogeneous product.

     The data from the EPA-Versar sulfur variability study(27) can be
used to show the reduction In both mean and variance of coal properties
by coal cleaning.  The study used the same sampling procedures, sample
preparation procedures, and laboratory analysis methods for both raw and
cleaned coal samples from R & F Coal Company and Republic Steel
preparation plants.  Figures 7-2 and 7-3 Illustrate the data.
     Studies have shown that any type of coal  handling operation
Introduces mixing of the coal which 1n turn decreases the variability.
Naturally occurring mixing takes place 1n crushers, silos, and holding
bins.  Block shuffling takes place whenever successive truckloads,
railroad loads, or silos are processed 1n a different order than the
order 1n which they were received.
7.3.4  Measuring Error Variability
     The total variance of the coal data may be split Into two pieces:
one due to the Inherent variability of the coal and the other due to the
                                   -81-

-------
oo
ro
J
A
L
s
u
c
R
               5 -
               3 —
               O ——•-
                   DHY BASIS
                                                 7/21 - 8/15
                      7/21 - 7/25     7/28 - 8/1
                  0
                       I
                     20
I
                                                8/5 - 8/8  -   8/11
                                                            I
                             I
      60       80          100
PRODUCTION  TIHE(HRS)
  I
1B0
                FIGURE 7-2.   REPUBLIC STEEL CORPORATION  - hourly Increment data  for total  sulfur.

-------
               ia
co
CO
        E
        II
        I
        s
        S
        I
        0
        N
L
B


S
0
a
/

ft
B
T
U
      8   —
                e
                     DRY IASIS
                                 Ron COAL
                          iv
                                          \^*^^~^V/^^
                                                                          CLCAN COAL
                      7/21 -  7/25
                              7/28 - 8/1
              8/5 - 8/8
                              1

                             20
                                                         +
8/11 - 8/15
I    '     f   ^    I     '     I

0        60       80         100


       PRODUCTION TIME(HRS)
                                                                          1

                                                                        180
    FIGURE 7-3.   REPUBLIC STEEL CORPORATION - hourly Increment data for sulfur dioxide emission rate.

-------
measuring process.  The measuring error variance may possibly be reduced
by better QA/QC procedures, more Increments per composite,  additional
composites, better equipment, more representative samples,  etc.   Alterna-
tively, CEM and Method 6B emission measurements have lower  error than
does CSA.  The Impact of measuring variance may be seen 1n  Figure 7-4.
7.4  AUTO-CORRELATION OF DAILY MEASUREMENTS
     Before discussing this Important parameter, a general  description of
correlation 1s provided.
7.4.1  Correlation
     A set of data, summarized by a measure of center and a measure of
spread, 1s far from complete.  We are usually Interested also 1n the
connection between the data.  Such a connection, or association, 1s
measured by the correlation coefficient.  Correlation 1s almost as common
as mean and variance 1n summaries of data.
     Correlation between two variables means a systematic connection
between changes 1n one variable and changes 1n the other.  When an
Increase 1n one variable tends to be accompanied by an Increase 1n the
other, the variables are positively correlated (associated).  When an
Increase 1n one variable tends to be accompanied by a decrease 1n the
other, the variables are negatively correlated.  For example the amount
of coal burned and the outside temperature are correlated.   In summer 1n
the  south, as the temperature goes up more electricity for air
conditioning 1s required.  Thus, the relationship between temperature and
the  amount of coal used are positively correlated.  Similarly, 1n winter
1n the north the relationship between the temperature and the amount of
coal used would be negatively correlated since the amount of coal used
Increases as the temperature decreases.
     It 1s often desirable to measure the correlation of two variables
relative to their standard deviations.  The correlation coefficient  "r"
1s this measure (covarlance of the two variables divided by the product
of their standard deviations).  The correlation coefficient 1s positive
when an Increase 1n the value of one variable 1s on the average matched
                                   -84-

-------
        RSD = Standard Deviation * mean value
00
01
                                    CEM ERROR

                                    RSD=.01
                                    OR
                                                              MEASURED
            INHERENT COAL

             RSD=.125
AAlA  -
                                    CSA ERROR
                                    RSD=.05
                                                             MEASURED
                              FIGURE 7-4. VARIABILITY OF DATA

-------
by a related Increase 1n the other variable.  It 1s negative when an
Increase 1n one variable 1s on the average matched by a decrease 1n the
other variable.                                                >!
     The correlation coefficient, which 1s often called the "correlation,"
always has a value between -1 and +1.  In particular, r = -1 means that
all of the data points fall exactly on a straight line having negative
slope; r = +1 means that all of the data points fall exactly on a
straight line with positive slope.  Correlations near either +1 or -1
Indicate that the points fall close to a straight line.  When r >0 the
scatter plot shows a trend from lower left to upper right and the line
about which the points cluster has positive slope.  For r < 0, the trend
1s from upper left to lower right and the slope 1s negative.
7.4.2  Auto-Correlation
     Auto-correlation occurs when the value of a single variable 1s
related to one or more values of the same variable.  The statistical
properties of a time series are measured by the autocorrelations between
observations.  An autocorrelation of order one 1s the correlation between
the time series observations which are separated by one time unit.  An
autocorrelation of order two 1s the correlation between the time series
observations separated by two time units*.   Similarly auto-correlations of
higher order can be defined.  All the auto-correlations are between -1
and +1.
     An auto-correlation of order one close to one Indicates that
observations separated by one time unit have a strong tendency to move
together 1n a linear fashion:  high values tend to be followed by high
values, low by low.
7.4.3  Auto-Correlation 1n Coal
     The geological formation of.a coal seam contributes to a spatial
structure continuity 1n the coal  mine.  This structural continuity 1s
somewhat preserved during mining, transport, and handling of coal and 1s
transformed Into a temporal continuity In  the coal entering the plant.
This Implies that any sample value 1s dependent on those of the
                                   -86-

-------
neighboring points (1n time) and 1s correlated 1n some fashion.  Strong
positive correlation Indicates that past data provides much useful
Information 1n predicting future events.  In other words, a higher sulfur
value at one point 1s more often followed by a high sulfur value at the
next point.
     Failure to account for the correlation structure 1n coal data
results 1n gross underestimation of emission violations of actual
coar   .  Consider the data given 1n Figures 7-5 and 7-6.  The mean,
the variance, the standard, and the averaging time for both data sets are
about the same.  Figure 7-5 shows the effect of the high correlation, 0.7
while Figure 7-6 shows the effect of a low correlation, 0.1.  In both
diagrams, the number of observations exceeding the standard are about the
same, but with a high correlation they group together while for a small
correlation they appear randomly.  Hence, the average emissions over the
"long-term" compliance period will not exceed the standard 1n Figure 7-6
while It clearly will 1n Figure 7-5.  Thus there-Is an exceedance of the
emissions standard for the data from Figure 7-5 while there 1s no
exceedance Indicated 1n Figure 7-6.
     In general, the larger the.correlation the harder 1t 1s to pass the
emissions standard.  (Blending 1s one effective procedure for lowering
the correlation.)
7.5  VARIANCE AND AVERAGING PERIOD
     For compliance coals leading to the same mean level  of sulfur
dioxide emissions, the longer the averaging period for compliance the
lesser the effect variance has on the likelihood of compliance.  Clearly
long averaging periods allow periods where the emissions  are below the
standard to balance those where they are above (see Figure 7-7).
Alternatively, the series of averages has a variance which decreases as
the averaging period gets longer (see Figure 7-7).
                                   -87-

-------
                                            EXCEEDANCE
SO,
                                                               STANDARD
                                                               AVERAGE
                               TIME
         FIGURE 7-5.  COAL DATA WITH  r  =  0.7.  EXCEEDANCES EXIST
                                    88

-------
so,
                                                                    STANDARD
                                                                    AVERAGE
                                 TIME
           FIGURE 7-6.  COAL DATA WITH r = 0.1.  NO EXCEEDANCES.
                                    89

-------
                          DAILY DATA
                                                                             STANDARD


                                                                             MEAN
                                                                    STANDARD


                                                                    MEAN
                       WEEKLY AVERAGES
                                                              STANDARD


                                                              MEAN
                     MONTHLY AVERAGES

  *T1me scales on weekly  and monthly averages expanded to preserve  same
   space relationship as  dally averages.
FIGURE 7-7.   EFFECT OF AVERAGING PERIOD OF  COMPLIANCE AND VARIANCE
                                     90

-------
                                SECTION 8
                                THE MODEL

    The practical application of a coal sampling and analysis (CSA)
program 1s Influenced by several factors.  Among these 1s the adoption of
a statistical theory that allows a small data set to represent a larger
population.  A certain amount of Information pertaining to the
population, or coal supply, must be known so that a small data set, or
the average of a small number of Increments taken 1n one day, can
represent the entire dally coal  supply.  The mean, variance, covaMance,
and correlation are useful parameters for describing a series of coal
characteristics.  This Information applied 1n an appropriate model  allows
an estimation, or forecast, of coal characteristics.
  .  This section presents a statistical model which Incorporates the
parameters of an auto-regressive model  of order one (AR(1))  and the
parameters of an auto-regressive moving average (ARMA(1, 1)) model  to
allow an estimation of coal characteristics.  The Input to this model 1s
based on measured data, for example,  Ibs SOp/MMBTU (calculated through
CSA).  The model would then estimate  the actual Ibs S02/MHBTU 1n the
coal supply.   The Input to the model  can also be based on measured  data
of the stack  emission; the output would be an estimate of the actual
stack emission 1n Ibs SO /MMBTU.  In  both cases the model provides  an
estimate for  the actual measure  of the  coal characteristic as opposed to
the measured  value.  The measured value (data) equals the actual value
plus the measurement error.  The measurement error consists  of
sampling/analytical errors.  The actual value of the SO- emission
parameter 1s  not constant as the coal property has some Inherent
variability.
                                    91

-------
    The focus of this section 1s on the application of a statistical
model to a CSA program.  Guidelines to minimize the measurement error are
discussed, Including considerations of sampling, sampling location, lot
size, and analysis.  In addition, 1t 1s shown how the model can be
applied to stack monitoring programs, specifically continuous emission
monitoring (CEH) and Method 6B (continuous bubbler (CB)).
    As a prelude to this, brief descriptions of the operations of CEM and
CB systems are presented.
8.1  EXPLANATION OF A CEM SYSTEM
    Continuous emission monitoring can be done by an extractive method or
an optical method.  In the extractive method a gas sample 1s extracted
from a sampling point 1n the stack or emission exhaust duct over a
specified time period.  The gas stream 1s sampled and analyzed
continuously except for short, periodic blow backs which clear the sample
line.  An SCL analyzer continually measures the S02 concentration In
the gas sample.
    The optical method of CEM uses a light source to measure SO-
levels.  The light source passes through the stack and the degree of
scattering of the light 1s read by a detector.  The S02 level 1n the
stack 1s calculated according to the relationship between the degree of
light scattering and SO  concentration.
    From this continual monitoring of the SO- level, hourly averages
which consist of four equally spaced data points per one-hour period are
generated.
8.2  EXPLANATION OF A METHOD SB/CONTINUOUS BUBBLER MONITORING SYSTEM
    Method 6B, Continuous Bubbler (CB) monitoring, 1s an alternative to
the Method 6 determination of S0_ emissions.  A gas sample 1s extracted
from a sampling point 1n the stack.   The CB monitoring system operates
continuously throughout the sampling period.  The SO  contained 1n the
sampled gas 1s separated and collected 1n the sampling train.  The CB
                                    92

-------
method composites the entire SO  fraction collected during the sampling
period.  The SO- fraction 1s measured by chemical analysis and the
emission rate for the sampling period can be determined.
8.3  DESCRIPTION OF THE MODEL
    The model simulates the coal properties present 1n either a moving
stream of coal or stack emissions.  In either case, the coal or stack gas
passing the sampling point can be thought of as being made up of Q
discrete packets of homogeneous material per day (Figure 8-1).  The
computer Implementation 1s based on the assumption that Q = 1,008,000
discrete packets pass the sampling point each 24-hour period.  When
applied to a CSA program the numerical results are based on the
assumption that the action of the sampler cutting through the coal stream
takes about two seconds.  Therefore, roughly 24 discrete packets would be
collected with each pass of the sampler.
    The actual value of the coal property 1n each discrete coal packet 1s
G.  For example G^ can be the total sulfur contained 1n the 1th
discrete packet of coal.  Each potential sampling unit 1s J.  Thus, for
coal sampling, every collection of 24 coal packets 1s a potential
sampling unit.
    The Increments, I, are the actual sampling units, which correspond to
the potential sampling units (J) that were actually collected.  The
collected Increments are composited to form C as shown 1n Figure 8-1.
    A slightly different approach 1s used when describing emissions
data.  By the time flue gases reach the stack they have become so well
mixed that each unit of gas has part of the products of combustion from
hundreds of the coal packets.  For a CEM system, hourly emission values
are based on 4 equally spaced data points.  Dally averages are the
average of the 24 hourly averages.  In terms of our model, N=96
Increments make up the dally composites (Figure 8-2).  We assume that a
sampling unit J 1s made each minute so M=15.  Hence, L=Q/MN=700 coal
units make up each sampling unit.
                                    93

-------
COAL FLOW
PAST
SAMPLING POINT





G,
i —


G2

• • •

GL



GL+1
	 it nuuna 	 —. 	

GU2

• • •

G2L


— "•. . X V. . . *J





C-^-J



GQ
L 	 }
\
                                                                                                     *2M
                                                                                                      \
                                                                                                      '2  +
                                                                                                        'NM
                                                                                                         \
                                                                                                   C = COMPOSITE OF INCREMENTS
                                                                                                                           J
vo
-p.
Q = LMN
Q = 1,008.000 discrete coal packets per 24 hour period.
L = # of coal packets collected in each potential increment (determined by coal sampler cutter speed).
M = difference between increment numbers for actual  sample increments.
N = # of increments actually collected to form a composite.
If L = 24. and N = 35 then: M = 1.008.0007(24 x 35) = 1200
                                 FIGURE 8-1. INCREMENT COLLECTION AND COMPOSITE IN A CSA PROGRAM

-------
&AS FLOW
PAST
SAMPLING POINT





G,
k»_—

fi2

• • •

fiL


• 	 >v 	 — '

°t+1

GL+2



• • •

G2L





L — „ — J


J2M



GQ
MM
                                                               JM

                                                               I
I
\
                                                                                     C = COMPOSITE OF INCREMENTS
In a CEM system each potential measurement, J, is   hundreds of discrete units G.
Every 15th  J. is selected to provide the increments, I.
The 96 I's are averaged to provide the daily composite, C.
Assuming Q = LNM = 1,008,000 we get:
L = 700 coal units measurement.
N = 96 increments per day, 4 per hour.
M = 15
                FIGURE 8-2.  INCREMENT COLLECTION AND COMPOSITE IN A CEM PROGRAM

-------
    The CB system 1s modeled 1n a similar manner.  Since the CB system 1s
continuously withdrawing well-mixed stack gases, 1t receives part of the
products of combustion from each coal packet. ' Thus, L=Q=1,008,000 and
M=N=1 (Figure 8-3).
    The variability associated with the coal property G 1s subject to
correlated and random elements.  This stochastic process 1s modeled by an
auto-regressive model of order one (I.e., an AR(1) process).
    Expressed 1n terms of our model, the AR(1) model relates the next
value of G to the current value of G using the constants v, the mean,
and , the auto-regressive parameter, via the equation:
where:
              11 + * (6t - p ) + a                        (8-1)
       G. = measurement at time t,
       G    = measurement at time t + 1,
       a. +, = normal random shock at time t + 1 ,
and
A random shock 1s a normally distributed random variable that Is
Independent of all past history.  It accounts for any Inherent variance
1n the data.  The auto-regressive parameter, $ , 1s a measure of the
correlation between the current data point G. and the next value
6 M '
    If u and 4> are known, then the AR(1) model Implies that the best
forecast for the next data point 1s the mean, v, plus a multiple of the
deviation of the present data point G. from the mean:
                     Best forecast = w + $ (G  - u ).
    It can be shown that the best forecast for the data point L units  1n
the future satisfies:
          Best forecast for G-^+L = u + L (current data - mean)
                               - ii + +L ( &t - v ) .
                                     96

-------
                                                               24 HOURS
  STACK GAS
  FLOW PAST
SAMPLING POINT
B,

G2
• • • • • • •
GQ
                                                                   C = DAILY COMPOSITE
             In a CB system each measurement contains part of each discrete unit G.
                              FIGURE 8-3.  INCREMENT COLLECTION AND COMPOSITE IN A CB PROGRAM

-------
 Hence  the  forecasts' dependence on the current data decays
 exponentially.   By  considering the relation  (8-1), 1t can be  shown  that
 no  future  value  1s  Independent of current values.  In fact, 1t can  be
 shown  that the correlation between any two data points L units apart 1s
 4>  .  Thus, for large time lags between data  points, the relationship
 1s  essentially one  of Independence but for small time lags 1t 1s a  highly
 dependent one.   Methods to estimate the parameters p and  from a
 data set will be discussed later.  First the description of the model 1n
 the terms of a CSA  program will be finished.
    Assume that  sampling 1s done periodically, J 1s the measured value of
 each potential sampling unit and Its value 1s  an average of L discrete
 coal packets (G) which make up each Increment.  Thus:
               * G2 .*
        J_
         L
L+I
   GU2 *
     JMN
      L+l
                   GMNL}
     Since the variability associated with the discrete coal packets 1s
modeled by an AR(1) model, J 1s modeled as* an average of an AR(1) model.
This latter time series model 1s termed an auto-regressive moving average
model, ARMA(l.l).  Its representative 1s
where
and
          t+1
               a (Jt-w)
measurement at time t
measurement at time
random shock at time t+1
random shock at time t
             -1  <
                                     e < * 1.
                                               - e
                                    98

-------
The future value of J depends not only on the current value of J, the
mean \t, the auto-regressive parameter o, but 1t also depends on the
shock to the system at time t and an appropriate weight e given to that
shock.
    If I,, I_, ..., I., are the actual sampled units, and Ju,
        1   2        N                                    H
Jou« •••• Jnu represent the coal property measured 1n the
 2M        Nn
corresponding Increment, then:
             IT  = JM
             I2 =
             TN " JNM-
Since I 1s a "thinned" version of J, 1t too 1s an ARMA(l.l) model.
Next, consider that one composite sample 1s formed each day from the N
Increments:
             C.Jd, *I2* ... *IM)

               = N (JH * J2M * ••' * JNH}-
The composites are^ then used to determine the quantities that are of
Interest, e.g., total sulfur (weight %), so that the actual measurements
1n the laboratory represent the composites (C) and not the potential
sampling units (J), nor the Increments (I), nor the packets (G).
    Again, C 1s an averaged value derived from the I values so C 1s also
modeled by an ARMA (1,1) model.
    In all cases the dally composite value, C, 1s the actual value of the
coal or stack emission.  It 1s provided by the model.  A CSA program
which composites the Increments and analyzes the dally composite would
calculate the measured value (X)  of the coal.  The measured value X would
not equal C because X Includes some unavoidable measurement error.  The
measurement error consists of variations 1n the sampling and analysis
procedures.
                                   99

-------
    The ARMA (1,1) model that describes the actual coal supply (C) can be
extended to a model for the measured value (X) of the dally composites.
Using the parameters that describe X and the properties of the model, the
parameters that describe the actual coal supply can be derived.
8.4  ESTIMATION OF THE MEASURED DAILY COMPOSITE
    To apply the ARMA (1,1) model which describes C there must be some
relation between Its parameters and the parameters that describe X.  The
value of X does not equal the actual value C because errors 1n the
measuring process will be reflected 1n the value X.  There are errors 1n
the sampling, sample preparation, and sample analysis that cause the
value X to be less representative of the actual coal.
    The relationship between C and X 1s:
                                X = C * ME
where ME 1s the measurement error of the sampling and analysis
procedures.  ME 1s modelled as a normally distributed random variable
that 1s Independent of all past history and of C.  The measurement error,
ME, and the actual value of the composites,.C, can be determined from the
model once the parameters that estimate the value of the measured dally
composites (X) are known.  Since C 1s an ARMA (1,1) model and ME are
Independent random shocks, X 1s also an ARMA (1,1) model.  The
auto-regressive parameter of the model for X 1s the same as the
auto-regressive parameter of the model for C.  The auto-regressive
parameter for X can be estimated from the data.  Thus, the
auto-regressive parameter for C can be estimated.  Based on this value
and the above model, estimates can be determined for all the parameters
describing the processes G, J, I, and C.
    The estimators needed to describe the dally composites, X, are (K 1s
the number of data points):
                                   100

-------
1.   estimate of mean =  A
                        v
                     3 average of the composites

                       1   K
                     S~           X
                            A
2.  estimate of variance = var  X
                             _        (X-X) 2                 (8-3)
                              A
3.  estimate of covarlance = cov.. X
                              ,    K-l
                             _  X   (X, - X) (X. -  - X)      (8-4)
                             K-l  1=1
                              A             -
4.  estimate of correlation = corr.  X

                               A     A
                            = cov^/var X                      (8-5)


5.  estimate of the second order covarlSnce =

A          ,  K-2
cov2 X =  _   I   (X,  - X) (X. 2 - X)                        (8-6)
          K-l   1=1


6.  estimate of the second order correlation =

         A
 A      cov X
corr2X =  2	                                               (8-7)

          var X
                               101

-------
    7.  estimate of the auto-regressive parameter,  AR(X)
                A
     A     cov, X
    AR (X) =/J _                                                (8-8)
              A
             cov1 X
    8.  estimate of the moving-average parameter MA(X):
                                                                   (8_9)
    where:       .
                 A
            C = corr.  (x)

                 A       A
            Y = AR(X)  - corr^X)

and t 1s used 1f Y 1s  negative,  - 1s used 1f Y 1s positive.
    The estimate of the auto-regressive parameter 1s the basic Input to
the model describing G.  It should be noted that equation (8-8)  provides
only a rough estimate  of the auto-regressive parameter.   A better
estimator may be obtained  via any Box-Jenkins Time Series routine.
    The parameters of  the  ARMA (1,1) model  which describe the dally
composites (X) can be  called a and B with a being the auto-regressive
parameter and 8 the moving average parameter.  An estimate for a 1s
provided by equation (8-8), the ratio of the second order covarlance to
the first order covarlance.  An estimate for B 1s also calculated through
the measured data via  (8-9).
    The parameters of  the  ARMA (1,1) model  which describes the actual
coal (C) can be called 
-------
    The relationship of  the  parameters «, B and f for the coal data X
and underlying coal  properties,  C, are given 1n the following two
propositions.   Their proofs  may  be found  1n (29).
Proposition 1.  Let  A be ARMA  (1,1) with  parameters « and B.  Let B be
the average of successive p  elements  from the series A, I.e.

                  Bk = p- o  1f and only  1f  B < f .
    In the above  results,  var (Y) means the variance of the series Y and
cov (Y) means the auto-covarlance of order one of the series Y.
    Now that the  parameters that describe the actual coal are known the actual
coal characteristic (e.g.,  Ibs  SO-/MMBTU) can be calculated.  The difference
between the B term 1n the  model describing the dally composites (X) and the
4 term 1n the actual  coal  (C)  model measures the splitting of the variance
of X Into Its two components:  the variance of the coal and the variance of the
measuring process.
                                  103

-------
8.5  EXAMPLE

     to Illustrate the use of equations 8-2 through 8-9, consider the
following data set.  A number of total sulfur determinations have been
made on K = 30 dally composite samples from train loads of coal
originating from the same mine.  The variability of the data set may be
attributed to the natural variability of the coal and to the measurement
error.  Me will split this value below Into Its two pieces.

     Day    Total Sulfur (wt.%)         Day    Total Sulfur (wt.%)
1
2
3
4
5
6
7
8
9
10
n
12
13
14
15
2.46
2.64
2.68
2.49
2.81
2.38
2.46
2.97
2.68
2.50
2.76
2.50
2.45
2.65
2.73
16
17
18
19
20
21
22
23
24
27
26
27
28
29
30
2.55
2.51
2.45
2.30
2.26
2.39
2.64
2.33
2.38
2.31
2.52
2.31
2.46
2.65
2.52
          Equation (8-2) gives the estimated mean of the data set

               K
               I   '    X.
                        .
              1=1
      =J_ (75.75) = 2.525.
       30
    The variance 1s calculated using equation (8-3)

     A              K         _ 2
    var X = _]_     I   (X,  - X)
            K-l     1=1     1
            J	  (.82115)  = .0283
            30-1

                                   104

-------
    Equation (8-4)  gives the estimated  first  order  covaMance  of  the
composites
           3 30Tl(°'1526)  = -00526-
    The first order correlation of the composites  1s  simply  the  first
order covarlance divided by the variance:
     A         A      A
    corr. X = cov1  X/var X

            = .00526/.0283
            = .185868.
    The second order covarlance 1s estimated  using equation  (8-6)

    C°72 X ' KIT   j ,    (XU2 - "
                   1=1

           = 2^- (.02545)

           = .0008776.
    The second order correlation  of  the composites 1s  the  second  order
covarlance divided  by the  variance
     A         A       A
    corr. X = cor_X / var  X

            = .03099.
    Finally, the ratio of  the covarlances  gives  the auto-regressive
parameter:
     A        A
    AR (X) = cov2 X
              A
             cov.  X

           = .16675.
                                  105

-------
    The next step 1s to split the total variance of the dally composites
 .0283 Into two parts: the variance of the underlying coal and the
 variance associated with the measuring process.  Since the
 auto-regressive parameter for both the true coal and the measured data 1s
 the same, we estimate the auto-regress1ve parameter for C to be 0.16675.
    In order to split the variance, we must compute the moving average
 term for the measured data X.  Since the correlation 0.1859 1s larger
 than the AR parameter 0.16675, we use (8-9) with a + sign to obtain
    A
    HA (X) = -.0198.
 Using this, we find:
    var (C) = .0145 = 51.3X of total  variance
    var (ME)  = .0138 = 48.7X of  total variance.
Hence, about half of the observed variation 1n the data 1s due to the
Inherent variability of the coal  while the other'half  1s  due to the
Imprecision 1n the measurement process.
                                   106

-------
                                SECTION 9
                   CONSIDERATIONS OF MEASUREMENT ERROR

    The previous section Indicated that the actual  characteristics of a
coal stream could be specified knowing the measured value and the
measurement error.  The measured value will be more representative of the
actual coal 1f the measurement errors 1n the sampling and analysis
procedures are minimized.
    This section discusses the considerations of sampling error, sampling
location, sample lot size, sampling system performance, sample
preparation, and sample analysis which contribute to the total measuring
process error.  Also, some elements of a quality control program which
may minimize the measurement error are discussed.
9.1  Sampling Error
    The sampling error 1s by far the largest component of the total
error.  Generally 80 percent of the total error 1s due to some aspect of
        CI7\                                   .
sampling*  '.  Sampling error consists of random error and systematic
error.  Random error 1s caused by some chance occurrence 1n the sampling
process.  The flow of coal to a sample cutter may vary so that Increments
of varying weight are Inadvertently collected.  Random error 1s also
related to the Inherent heterogeneity of coal.  Successive Increments may
exhibit great differences 1n size or chemical composition.  Random error
generally occurs 1n both positive and negative directions around the mean
and tends to have an average value of zero.
    Systematic error, also called sampling bias, 1s an error 1n
representativeness usually caused by Inaccuracies 1n the sampling
mechanism.  For Instance, the geometry of a sample cutter passing through
a stream may consistently cause the larger particles to be omitted from
                                   107

-------
the collected sample.  Another example 1s a sample cutter whose
dimensions are not large enough for the entire width of a stream.   As a
result a portion of the coal stream Is consistently omitted.   Both of
these examples Illustrate the non-random bias due to sampling.  Because
bias 1s a repetitive error, as opposed to random, the results tend to be
consistently higher (or consistently lower) than results from an unbiased
sample.
    In a CSA program an Increase 1n variance caused by random error can
mean a broader range of data around the mean emission, and therefore a
greater probability of exceedances.  There are several ways to minimize
the potential Increased variance due to random error.  The ASTM approved
sampling methods and techniques discussed 1n Section 3 are designed to
decrease random error.  For Instance, a sampling method that  takes a
relatively large number of Increments lessens the random error.  Examples
from charts will follow.
    When the systematic error 1s considered, the bias present 1n the
sampling mechanism must be known.  A widely recommended test  for bias Is
a statistical comparison of the data from coal collected by the automatic
mechanism and of the data from coal collected by the ASTM stop-belt
technique   .  Statistical tests may determine whether a bias exists
and the significance of the bias.  Considerations of sampler  bias  are
further examined 1n Section 10.
9.2  Sampling Location Considerations
    The goal of a CSA program to predict SO  emissions should be to
attain the most accurate relation between the coal sampled and the
resultant SO- emissions from that coal.  It 1s Important that the
actual stack emissions can be directly related to the coal data from a
specific time period.  Sampling coal feedstock from an as-fired location
Immediately prior to Us entering the boiler ensures that the stack
emission 1s a result of that lot of coal.
                                        108

-------
    Another aspect of the as-fired location which must be considered 1s
the number of feed lines to the boiler.  Withdrawing a sample from one of
many  feed line may not accurately represent the total coal flow Into the
boiler.  Sampling from every feed line will reduce the uncertainty of the
coal data.
    Sampling from locations further from the boiler results In a need for
lag time calculations to predict when the lot of sampled coal enters the
boiler.  For example, when coal 1s sampled as 1t 1s delivered to the
storage bunkers, the lag time 1s a function of the bunker capacity,
feeder operation, and feed rate to the boilers.  The lag time calculation
assumes that the lot of coal represented by the sample moves as an
unchanging mass toward the boiler.  Instead several variables may affect
the flow characteristics of the coal after 1t has been sampled and may
subsequently affect how well the sampled coal represents the predicted
SO- emissions.
    Coal processing and handling devices such as pulverizers or storage
bunkers change the flow characteristics so that samples taken upstream of
these devices tend to be only an approximation of the flow character-
istics of the coal entering the boiler.  The coal undergoes diffusive
mixing where adjacent lots of coal are handled together, and block
shuffling where the sequence 1n which lots of coal are handled 1s
changed.  As a result, the predicted SO. emissions for a certain time
period may be based on a lot of coal that does not enter the boiler
during that time period.  Also, the coal may be so thoroughly mixed with
other lots of coal that the coal data obtained at the upstream sampling
point 1s meaningless.
    The as-received sampling point 1s even more affected by lag time
problems.  In the case of rail shipments, samples may be withdrawn from
each rail car as 1t 1s emptied Into the unloading hopper, or may be
withdrawn from the conveyor belt below the hopper.  A CSA. program that
relies on coal data from samples taken as-received must take Into account
the fact that a tralnload of coal 1s generally divided Into a supply for
plant use, and another portion which Is diverted to the storage pile or
                                      109

-------
coal reserve containing as much as a 91-day supply of feedstock.   Coal  1s
reclaimed from the storage pile for nights, weekends, and plant operating
periods when no coal shipments are received.  The as-received Incremental
samples may be collected from several cars 1n a tralnload and the
composite used to generate the dally coal  data.  Considering that the
dally tralnloads come from the same mine,  1t 1s possible that the mean,
variance, and covaMance are relatively uniform.  In that case an
estimate can be made of the quality of coal going Into the plant.  This
estimate will be subject to the uncertainties and assumptions that are
present when sampling from a lag time location.
    The dally coal data can also be used to estimate the coal properties
of the storage pile when 1t 1s reclaimed for plant use.  However, the
storage pile may contain coal from many tralnloads so that the
consideration of more than one dally coal  data set would prove more
representative.
    Because of the general uncertainties associated with sampling from
locations that require a lag time calculation, there 1s less assurance
that the predicted S02 emission can be accurately related to the
specific lot of coal sampled.
9.3  Averaging Time and Lot Size

    Another aspect of a CSA program that must be considered 1s the
averaging time used to determine compliance with SCL emission
standards.  Averaging time may be monthly, dally, or as short as three
hours.
    A shorter averaging time results In a  smaller lot size.  Emissions
based on smaller lot sizes have a higher variance than data from larger
         (27)
lot sizes    .  The higher variance could  lead to more exceedances of
the emission limit than would be seen with the same coal over a longer
averaging time.  The short-term effect of  the variance of coal properties
                                   (21}
can cause higher peak sulfur valuesv  '.  These high SO- emissions
may be high enough or long enough to cause the average SO  emission
over a short time period to be above compliance levels.
                                       110

-------
9.4  CSA Quality Assurance/Quality Control
    The purpose of a CSA program 1s to obtain samples which are
representative of the lot of coal from which the samples were taken.  The
coal data from the analysis of the samples Is used to estimate SO
emissions.  To ensure high quality results, a quality assurance/quality
control (QA/QC) program must be followed.  A quality control program
would Include the routine checks that are a part of standard procedures,
for example, duplicate analyses, split samples, and periodic calibrations.
    A quality assurance program, Intended as a means of evaluating the
effect of the routine procedural checks of quality control, Includes
Interlaboratory comparisons of results and Independent surveys of the
performance of standard methods.
    The following considerations for ensuring that the entire CSA program
1s producing high quality results will be presented 1n two areas: (1)
sampling and (2) sample preparation and analysis.
9.4.1  Sampling
    The major factor that must be considered 1n the QA/QC evaluation of
the sampling portion of a CSA program 1s the performance of the sampling
system.  An automatic sampler reduces the factor of human discretion 1n
sampling but does not eliminate 1t completely.  Faulty maintenance or
adjustment errors have a major effect on the final result.  A maintenance
program must ensure that the sampler 1s running properly.  Records should
be kept of the sampler's operation so that any system failures and down
times can be noted.
    Sampling personnel, besides ensuring that the sampling system 1s
operating according to standard procedures, should also document each
step of the sampling procedure by maintaining complete and accurate
records.  To completely document the sampling system's performance,
Information such as coal feed rates, sample cutter speed, weight of
sample Increment, number of Increments per averaging period, width of
sample cutter, and total lot size represented should be recorded.
                                      Ill

-------
    In addition, the boiler operating parameters should be documented.
The boiler firing rate, percentage of excess air, and generating load are
all Important parameters and should be recorded.  In addition, any
periods of operational problems or down time should be documented.
    The operating parameters of any SO. pollution control devices may
be Included 1n the recording of plant operation.  Any device which has an
effect on the emission level of S0_ should be Included, such as
flue-gas desulfurlzatlon units or partlculate control units.  A
partlculate control unit such as an electrostatic predpltator or a
mechanical dust collector will reduce SO  emissions by removing SO
which may be collected on the surface of an ash particle.  Any equipment
breakdowns or malfunctions that result 1n a pollution control system
being bypassed should be documented.
    In addition to these quality control procedures, an Independent
observer not associated with the normal sampling operation may review the
sampling procedure to assure that standard methods are being followed.
9.4.2  Sample Preparation and Analysis
    An Important QC procedure for sample preparation 1s periodic sieve
tests on the product of the sample crusher.  This allows an evaluation of
the performance of the equipment and the procedures.  Records should be
kept, documenting each step of division and reduction 1n sample
preparation.
    The analysis of duplicate samples 1s recommended as a QC of the
analytical procedures.  Additionally, all analytical equipment should be
periodically Inspected and calibrated.  During the analysis, accurate
records should describe each step.  Standard methods and procedures
should be followed.
    The method of division, reduction, and analysis must be consistent so
that any large difference In analytical results can be attributed to the
sample, and not to an error 1n preparation or analysis.
    Quality assurance measures may Include reprodudblHty tests, where
split samples are analyzed by different laboratories using the same
standard method, and the results compared.

                                      112

-------
                                SECTION 10
                      CONSIDERATIONS OF SAMPLER BIAS

    A sampling system with a systematic error, or bias, will consistently
misrepresent the actual coal 1n the same way.  This can lead to
Inaccurate analytical results.  To prevent this a sampling system can be
bias tested to determine 1f 1t 1s Introducing any systematic errors.  Of
course, random sampling errors will always be present.
    To test for bias the sample collected by the sampling system must be
compared to a reference or standard sample.  A very accurate reference
sample 1s a full cross-section cut from a stopped belt.  The reference
sample Increments and the test sample Increments are paired observations
which are then tested for systematic errors.
    The method of testing for systematic error applies statistical tests
of significance to the paired Increments representing the same material.
The statistical tests of significance test the hypothesis that the mean
of the differences between reference and sampled pairs 1s zero.  The
results are Interpreted as either the absence or presence of bias at a
level of practical significance.
10.1  Bias Test Procedures
    The American Society of Testing and Materials has published a
Proposed Test Method for B1as^   .  In general, this method 1s the most
detailed and controlled procedure for testing the performance of
automatic sampling systems.  The proposed method describes the procedures
for testing for systematic error between paired observations.  The
recommended reference sample 1s a full cross-section cut from a stopped
belt.  This reference sample would be considered the best approach for an
absolute bias determination.  However, secondary reference samples
allowed by the ASTM when stopped belt Increments cannot be collected
would provide only a relative bias determination.
                                   113

-------
    It 1s Important to ensure that each reference sample Increment and
Its paired test sample are the same weight and are prepared and analyzed
under the same conditions.  In the case of two or three stage sampling
systems the primary test Increment 1s reduced to a crushed subsample.
Manual techniques of division and reduction will have to be applied to
the reference sample.  It 1s possible that the test sample may require
less preparation for analysis.  This 1s a possible source of bias.  It 1s
also Important to note that 1f a mechanical system has a bias then 1t 1s
better to correct the sampling system than to compensate for 1t
mathematically.
    When conducting the final statistical analysis 1t 1s not clear, as
stated by the ASTM, that the procedure of randomly excluding the excess
paired observations "assures that the boundaries of practical Importance
and statistical significance coincide."
10.2  Size Bias of Automatic Samplers
    The Kaiser Engineers of Pennsylvania, Inc. conducted bias tests on a
Run-Of-H1ne (ROM) raw coal sampler at an EPRI Coal Cleaning Test
        (31)
Facility    .  The bias test was performed following procedures similar
to the ASTM Proposed Test Method for Bias*30*.  The bias test samples
were collected by an automatic plate type sampler.  The reference sample
was collected by the stopped belt method.
    The results of the Kaiser test showed a bias 1n the size distribution
collected by the test sampler (Table 10-1).  A sampler's size
distribution bias 1s determined by measuring the percentage of each size
fraction actually collected and comparing 1t to the size distribution of
a reference sample.
    The &PRI sampler appears to collect less of the larger sized coal
particles (1.5 Inch and larger), more of the medium sized particles (8
mesh -3/4 Inch), and less of the smaller sized particles (less than 24
mesh) (Table 10-1).  That 1s, a sample collected by the EPRI sampler Is
less representative of the larger and smaller particles than the
reference sample.  The EPRI sampler 1s biased relative to size
distribution because 1t does not collect all of the fine particles and
larger particles present 1n the actual coal stream.
                                   114

-------
   TABLE 10-1.  SIZE DISTRIBUTION OF EPRI BIAS TEST DATA.
REFERENCE SAMPLE
EPRI TEST SAMPLE
COAL SIZE
FRACTION
3" x 1 1/2"
1 1/2" x 1"
1" x 3/4"
3/4" x 1/2"
1/2" x 1/4"
1/4" x 8 mesh
8 mesh x 28 mesh
28 mesh x 48 mesh
48 mesh x 100 mesh
100 mesh x 150 mesh
150 mesh x 325 mesh
325 mesh x 0
t OF WEIGHT IN
SIZE FRACTION (t)
15.07
16.34
12.46
13.64
14.82
9.52
7.01
2.34
1.59
0.60
1.20
3.74
CUMULATIVE
DISTRIBUTION (t)
15.07
31.41
43.87
57.51
72.33
81.85
88.86
91.20
92.79
93.39
94.59
98.33
% OF WEIGHT IN
SIZE FRACTION (%)
12.82
16.76
13.38
14.84
16.20
9.49
6.72
2.24
1.42
0.55
1.04
3.63
CUMULATIVE
DISTRIBUTION (I)
12.82
29.58
42.96
57.80
74.00
83.49
90.21
92.45
93.87
94.42
95.46
99.09
RATIO OF REFERENCE
TO TEST SAMPLER
0.8507
1.0257
1.0738
1.0880
1.0431
0.9968
0.9586
0.9573
0.8931
0.9167
0.8667
0.9706

-------
10.3  Sulfur and Ash Bias Simulations
    Using actual coal characteristics the results of Section 10.2 may be
translated Into possible sulfur or ash biases for the EPRI sampler.  For
this simulation two representative coals were tested:  Helvetia coal mine
(supplying Homer City) and Lower Freeport coal, Harrison County, Ohio.
For each, the sulfur, ash content, and relative weight was known for each
coal size class.  The relative weight for each coal  size collected by the
sampler can be found by using the bias figures 1n Table 10-1.   These can
be converted to sulfur and ash content for the collected coal.   These
latter can be compared to the actual  coal to determine the sulfur and ash
bias.
    In more detail, consider Table 10-2.  The coal  1n 3" x 1  1/2" size
fraction contains 6.99% ash, 1.27% sulfur and represents 54.4  pounds of
each 100 pounds of coal.  Thus of every 100 pounds  of coal, this size
fraction contributes 3.803 pounds of  ash and .691 pounds of sulfur.
Performing this same calculation for  each size fraction shows  that each
100 pounds of coal contains 7.165 pounds of ash and  1.465 pounds of
sulfur.
    Next suppose the EPRI sampler was used to sample this coal  and Us
operating characteristics for this coal were exactly as reported 1n Table
10-1.  Then for every 100 pounds of coal collected  by the sampler only
49.39 pounds (Instead of 54.4) would  be 1n the size  fraction 3" x 1 1/2"
since the sampler under-collects 1n this size fraction.  This  would
contribute 3.452 pounds of ash and .627 pounds of sulfur.  Performing
this same calculation for each size fraction shows  that each 100 pounds
of collected coal contains 7.167 pounds of ash and  1.484 pounds of
sulfur.  Hence the relative bias 1n sulfur 1s
                     (1.484-1.465)71.465 = 1.3% high
and the relative bias 1n ash 1s
                    (7.167-7.165)77.165 = 0.03% high.
    Repeating the calculations for the Lower Freeport coal (see
Table 10-3) produced relative biases  of - 0.6% low  for sulfur  and * 0.2%
high for ash.
                                   116

-------
Table 10-2.  HELVETIA COAL NINE DATA
Coal Size
3" x 1 1/2"
1 1/2" x 1"
1" x 3/4"
3/4" x 1/2"
1/2" x 1/4"
1/4" x 8 mesh
8 mesh x 28 mesh
28 mesh x 48 mesh
48 mesh x 100 mesh
i
100 mesh x 150 mesh
150 mesh x 325 mesh
325 mesh minus
Ash
Content
6.99
7.14
7.06
7.16
6.85
6.77
7.80
10.25
10.82
11.47
12.86
14.24
Sulfur
Content
1.27
1.52
1.54
1.72
1.73
1.82
1.84
1.83
2.13
2.36
2.58
2.74

Weight
(per 100 Ibs)
54.4
10.9
6.2
7.7
8.6
4.8
4.3
1.6
0.6
0.2
0.4
0.3
Actual Coal
Ash
(per 100 Ibs)
3.803
.778
.438
.551
.589
.325
.335
.164
.065
.023
.051
^043
7.165
Collected Coal
S
(per 100 Ibs)
.691
.166
.095
.132
.148
.087
.079
.029
.013
,.005
.010
1.465
Sample
Bias
0.8507
1.0257
1.0738
1.0880
1.0931
0.9968
0.9586
0.9573
0.8931
0.9167
0.8667
0.9706
Relative
percent
46.28
11.18
6.66
8.38
9.40
4.78
4.12
1.53
0.54
0.18
0.35
0.29
93.69
Weight
(per 100 Ibs)
49.39
11.93
7.11
8.94
10.04
5.11
4.40
1.63
0.57
0.20
0.37
0.31
Ash
(per 100 Ibs)
3.452
.852
.502
.640
.688
.346
.343
.167
.062
.023
.048
.044
7.167
S
(per 100 Ibs)
.627
.181
.110
.154
.174
.093
.081
.030
.012
.005
.009
.008
1.484

-------
                                                              TABLE  10-3.  LOWER FREEPORT COAL DATA
  Coal Size
   1  1/2" x  1"
   1" x 3/4"
  3/4" x 1/2"
   1/2" x 3/8"
  3/8" x 1/4"
^ 1/4" x 7 mesh
oo 7 mesh x  14 mesh
   14 mesh x 28 mesh
  28 mesh x 48 mesh
  48 mesh x 100 mesh
   100 mesh x 200 mesh
  200 mesh minus
Ash
Content
t
15.3
13.0
13.6
11.0
12.5
11.4

10.6
11.2
11.4
12.3
15.6

19.9

Sulfur
Content
%
2.52
2.67
2.89
3.04
3.30
3.38

3.22
3.23
3.12
3.65
5.54

5.38


Weight
(per 100 Ibs)
14.6
9.7
16.4
11.8
8.9
14.3

8.3
5.5
2.3
4.1
1.4

2.7

Actual Coal
Ash
(per 100 Ibs)
2.24
1.26
2.23
1.30
1.11
1.63

0.88
0.62
0.26
0.50
0.22

0.54
12.78
Collected Coal
S
(per 100 Ibs)
.368
.259
.474
.359
.294
.483

.267
.178
.072
'.150
.078

.145
3.126
Sample
Bias
1.0257
1.0738
1.0880
1.0431
1.0431
0.99681
2
0.9586
0.95862
0.9573
0.8931
0.90003
3
0.9515

Relative
percent
14.975
10.416
17.843
12.309
9.284
14.254

7.956
5.272
2.202
3.662
1.260

2.569
102.002
Weight
(per 100 Ibs)
14.681
10.211
17.493
12.067
9.102
13.974

7.800
5.169
2.159
3.590
1.235

2.519

Ash
(per 100 Ibs)
2.25
1.33
2.38
1.33
1.14
1.59

0.83
0.58
0.25
0.44
0.19

0.50
12.80
S
(per 100 Ibs)
.370
.273
.106
.367
.300
.472

.251
.167
.067
.131
.068

.136
3.108
   'bias used for 1/4" x 8 mesh coat
  2bias used for 8 mesh x 28 mesh coal
  ^biases calculated on assumption that one-third of the coal in  150 mesh x 325 mesh was in the size fraction 150 mesh x 200 mesh.

-------
     These results show that bias tests for size distribution must be
distinguished from bias tests for coal properties like sulfur and ash.  In
this limited example, using one estimation of a sampler size distribution bias
and two sets of coal data, the size distribution bias Introduced by the
sampler did not bias the sample for sulfur or ash.   Essentially, the EPRI
sampler collected too few of the finer coal particles and of the larger coal
particles.  As these sizes represent extremes 1n both sulfur and ash content,
undersampUng them both had a compensation effect.
10.4  Theoretical Calculations
     A more theoretical model for sulfur and ash bias can be constructed
using the Rosin and Rammler equation:
            n
          mX
        be                                                         (10-1)
where
       R = weight percent retained on a sieve of opening size X,
       X = sieve size opening,
    m, b = constants,
       n = size distribution constant.
    To adapt this equation to a bias test,  the size distribution of the
reference sample 1s fitted to an exponential  curve (Figure 10-1).   Using
the EPRI data 1n the bias test for the run-of-m1ne automatic sampler
(Table 10-1), the percent retained,  R, and  the size of sieve opening 1n
millimeters, X, are fitted to an exponential  curve to obtain:

   R . 97.3615-°-°46X.                                              (10-2)
Solving for X yields

    X = (-1) In (R/97.3615).                                       (10-3)
        0.046
                                   119

-------
                                                         KMCfN  OPCMNO,
                                                                      1*1
po
o
i
s
1
10
IS
t«
JO
48
PERCENTAGE BY WOGMT
: SS s a s 8 2 ?
Ul •'
i S?
0 »2
V tl
t9
tl
1-H

tt
a t








--. .















x'


« .OS








. -
















( )

f



MO7

-






.. -.
















- —





400 123 270 2
1 1
400 S3 170 2C
US
o»


























' '




Ml
10
J!
























/
r">






























f.
















.













'"








—





















^




—


































1






















Br- —






























x
































-4












-









-







r










.6
-

















-










10




















^











—














































^
--































X*





























X

/































)L












MO WO 10 1 M tO 4O JO 2O 1* M 14 12 10 • _«,
US STANDARD SIEVE DESIGNATION
III I'll 1 1 1 1
ISO UO 65 4» 99 2* 10 It 14 1! K> * •
TYLER SIEVE DESIGNATION

—















^
—



























X*


•^


























•




—












• 4 -.
• 4
1
•-'
-






-














—

1
-








^



'


-












*
M
10
2







I
,

^?

1
c
r
i
















-»










— If
f
/



















*i
n
O








i
^






















*
!
SCR
0 4






^

























EEN
) 0/3 ID
































»"**(s —
— *
2
OPENINI
•o n
to fto




**





- - -

















. - ..


j MCHI
9
21
































1,
E3

jT






























1
tl
/































i
0 »j
































1
































o
1
s
7
K>
S
20
28
90
45
(0
68
70
75
•O
• 2
e»
•7
SO









t
                                                                                         (k)S«wi»
                                                                                                       cwllMelloi WIlM Mlbi.
                                                                                                                         Son.pl. tinft*y*--""«T..t No	


                                                                                                                         Maltrlof.	
                      GRAPHICAL FORM FOR REPRESENTING DISTRIBUTION OF  SIZE  OF BROKEN  COAL


                                       FIGURE 10-1.   Size Distribution of  Reference Sample.
                                                                                                                         Sl*v« Anolyllt by-

                                                                                                                              Dal*	
                                                                                                                         Plotlti)
                                                                                                                              Dal*.
                                                                                                                          Wtlghl ol gioti iomplt._


                                                                                                                          W«l«hl of Ittt •ompl«:_
                                                                                                                          Six* Anatyiii:
Paint I Sl*«*

_Np_ lOBiolng.
                                                                                                                              M
                                                                                                                                           Ett£
-------
    To compare the results from the reference sample data to the EPRI
sampler the same equations are used.  Figure 10-2 Illustrates the
relationship between the weight percent retained by the sampler and the
sieve opening:

    R1 . 100.0537e-°-04995 X.                                      (10-4)
Solving for X yields,

    X = (-1) In (R}l00.0537)                                       (10-5)
        0.04995

    Combining the calculated X value 1n each case (Equations 10-3 and
10-5) gives:

    (-  1  \ln/  R    \= (-  1     \lnf  R1    "\   .              (10-6)
    ^Q.046/    ^97.3615j  V 0.04995J  (.100.0537 /

Solving for R , the weight percent retained derived from the EPRI
samplers data, yields:

    R1 = 100.0537 /   R   N1-0859                                  (io_7)
                  \97.3615/
This equation calculates for each size fraction the percentage of the
reference sample that Is actually being collected by the EPRI sampler.
    The sulfur content, S, 1n each size fraction of the reference sample,
R, can often be expressed by:
    R = a In S * b                                                 (10-8)
where a and b are sulfur distribution constants.  Similarly for the EPRI
sample the sulfur content 1s expressed by:

    R1 = a In S1 + b.                                              (10-9)
Therefore, substituting equations 10-8 and  10-9 Into equation 10-7 gives:

    a In S1* b = 100.0537 /a In S * b V'0859                     (10-10)
/a In S * b V
^ 97.3615   J
                                   121

-------
                                                      1C ft (CM  OFENMO
ro
I
1
$
1
f
to
IS
so
IB
10
49
ISIZE , PERCENTAGE BY WEIGHT
5 5 8 58 3S2SSS
e i\
0 »2
' *4
•9
• 1


a i
— •





-'. '.'



















4 H









....















()"
.1


)(










-















-



«OT

-







~"
-

...



—






^





JO*
I


























• *





1
»

-
























'































i 	
' S-































*f~































&




















	





•




•







-

















—




-f







1























^
' ^








. \






















'
-^

































'












-








-






s
^










IO




















. ^

'"




























-*
,






























--
-^





























^







•







-.
















s

A





























s

f












400923270 200 MO KM (0 1 4O 80 40 SO 2O li » 14 12 10 i .
^I.S. STANDARD SIEVE DESIGNATION **
II II 1 ' II III
400S9tTO 200 HO WO 65 4S B M to 1* 14 II 10 • •
TYLER SIEVE DESIGNATION

--
















-
-



>



























/














\

















'














• 4
1
~










-


















9
-













/


-













IO
)







|
,




t
/
V
1
i















** —P










— ^
'/
/



















•>
1
n
O








£
























*
SCf
0 4.O 801





C
f
S
_^
























;
IEEN OPEN
•O «M
DTO »O





'
• '


























M *> 1
NG MCH
>
tl




s




























','



t





























i
tl


^
*





























•
0 t





—




























































0
1
a
r
10
8
20
29
JO
8-
49
90
M
CO
TO
79
eo
«2
•a
•T
•0
91
S2
n
t4
95
M
S9


1
                                                                                     (<)A«y mlt.M M •«•• l»f«. ahcM4i Mitt to»M «•!•
                                                                                           Ml •»»«< (I CMtbiMllM WliM Mfllt.
                                                                                                                   Sompl* No.
I No	r	
                    GRAPHICAL FORM FOR REPRESENTING DISTRIBUTION OF SIZE  OF  BROKEN COAL
                                                                                                                   Sl**« Anoljtlt b»-
                                                                                                                        Oalt—_
                                                                                                                   nalltd b>:_
                                                                                                                        Dolt.
                                                                                                                    Wtlghl al grati torapU:.
                                                                                                                    Wtlflht al lt
                                                                                                                    Si»v«  Anoly»ii:
                                                                                                                      Point  Slt«i
                                                                                                                      JE.
                                                                                                                                  itaintii
                                                                                                                                 s
%t.
                                                                                                                                 lL.A.6
                                                                                                                               r±£t
                                                                                                                                  J-dV
                                                                                                                                       Cumutj|i»i
                                                                                                                                        / J- S'd
                                                                                                                                       £!•
                                                                                                                                         i^:.i
                                                                                                                    Chmcltfiilict:
Point
A
B

Ptrctntog*


Difltune*
X,inch««



Y, inchit



                                                                                                                    Olilributlon
                                                                                                                    AtKOluU ill! militant 1 jj.  /H
                                        FIGURE 10-2.  Size Distribution of Test Sample.

-------
Solving for S  will express the sulfur content of the EPRI sampled
weight fraction 1n terms of the reference sampled weight fraction:

    S]= exp   M 00.0537  fa In S + b V'0899 . b \               (10-11)
              ^   a      I.   97.3615  /         a )
     To Illustrate the use of equation (10-11) and the effect of sampler
bias on sulfur content determinations the following  data 1s provided
(Table 10-4).  This table presents the sulfur content found 1n various
coal size fractions of a full cross-section cut of a stopped belt.  This
sample 1s the reference sample.  Using this data and equation 10-11 which
estimates the effect of the samplers bias on the total sulfur
determination, the test sample, S , can be compared  to the reference
sample, S.
    The first step 1s to fit the total sulfur content of the reference
sample, S, and the corresponding size fraction, R, to a line of
reasonable fit (Figure 10-3, correlation coefficient = 0.89) using a
logarithmic relationship:
    R = 93.042 In S * 33.253.                                      (10-12)
Substituting this expression describing the relationship between total
sulfur and size fraction 1n the reference sample Into equation (10-11)
and simplifying gives:

    S1 = exp (1.0754 (0.9956 In S f 0.3415)1'°859 -  0.3574).      . (10-13)
This equation Indicates the sulfur value that would  result from samples
collected by the biased sampler.  A range of theoretical reference sample
sulfur values and the calculated sulfur values collected by the biased
test sampler are presented 1n Table 10-5.  This table Illustrates how the
amount of sulfur 1n the reference sample affects the degree of bias, and
therefore the amount of sulfur, 1n the test sample.   The bias, expressed
as percent error, 1s more pronounced when the sulfur content 1s highest,
at 2.8 percent.  This 1s due to the fact that the finer particles of coal
have a higher sulfur content (weight percent)  than larger particles
                                   123

-------
TABLE 10-4.   SIZE FRACTION VS.  SULFUR AND  ASH  CONTENT.
COAL SIZE
FRACTION
>2"
2" - 1 1/4"
1 1/4" _ 3/4"
3/4" - 3/8"
3/8" - 4 mesh
4 mesh - 8 mesh
8 mesh - 14 mesh
14 mesh - 20 mesh
20 mesh - 28 mesh
28 mesh - 35 mesh
35 mesh - 48 mesh
48 mesh - 65 mesh
65 mesh - 100 mesh
100 mesh - 200 mesh
<200 mesh
SIZE
FRACTION
(WEIGHT %)
45.0
14.1
12.4
11.5
4.8
4.8
2.8
1.1
0.8
0.7
0.5
0.3
0.3
0.4
0.5
SULFUR
CONCENTRATION
(WEIGHT '/.)
1.22
1.50
1.54
1.72
1.73
1.82
1.85
1.84
1.79.
1.80
1.90
2.04
2.22
2.36
2.74
ASH
CONCENTRAT
(WEIGHT %)
6.94
7.24
7.06
7.16
6.61
6.77
7.71
8.31
8.86
10.28
11.24
10.84
10.79
11.47
14.24
                             124

-------
              100-
                                                                      0
                                                                              0
                    0
                                                                                                ©
                                                  0
00
Ul
              90-
              80-
              70-
              60-
              50-
                                                 0
                                              0
                                                            0


                                                           0
                                                                 0
                              0
SOLUTIONS FOB R ° F (S)                  CORRELATION

NON LINEAR MULTIPLE REGRESSION
 £n R = - fin 5.022 + 50.48 fin S + .037 fin S2          0.75

POWER CURVE
 Co R = fin 40.451+ 1.2847 fin S                    0.88

LINEAR
 a. FIRST 8 POINTS
    R = 84.389 S- 60.691                        0.09
 b. LAST 6 POINTS
    R = 3.137 S + 92.201                         0.92
 c. FOURTEEN POINTS
    R = 50.515-4.095                          0.85

LOGARITHMIC
 a. FOURTEEN POINTS
    R = 93.042 So S + 33.253                     0.89 -
 b. TWELVE POINTS
   1 R = 92.761 fin S + 31.755                     0.92

EXPONENTIAL
 H = 24.438 e'691 S                             0.83
                        I     I     I     r    I      I      I     I     I     I     I
                       1.1    1.2    1.3   1.4   1.5    1.6    1.7    1.8    1.9   2.0    2.1
                                                           S WEIGHT PERCENT
                   2.2   2.3
 I
2.4
2.5
 I
2.6
2.7    2.8
                                            FIGURE 10-3.   SIZE FRACTION  VS.  SULFUR CONTENT

-------
              TABLE 10-5.  BIASED SAMPLER SULFUR CONTENT
  REFERENCE SAMPLE             TEST SAMPLE
SULFUR CONTENT (WT V.)       SULFUR CONTENT (WT %)      PERCENT ERROR
1.0
1.1
1.2
1.3
1.4
1.5
. . 1.6
1.7
1.8
1.9
2.0
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
0.977
1.078
1.181
1.286
1.392
1.500
1.609
1.720
1.832
1.945
2.059
2.175
2.292
2.409
2.528
2.647
2.768
2.891
3.013
2.2
2.0
1.6
1.1
0.6
0.0
-0.6
-1.2
-1.8
-2.4
-3.0
-3.6
-4.2
-4.8
-5.3
-5.9
-6.5
-7.0
-7.6
                                    126

-------
                                                          B = 0%
                                                          AREA = .79
                                                          ft- 5%
                                                          AREA = .SO
                        -.0475 K
                                                          8 = 10%
                                                          AREA » .21
                 -.095 pt   -.0475 M
                                                          8=15%
                                                          AREA = .05
-.1425 n
-.0475 n
          FIGURE  10-4.  AFFECT OF  BIAS ON  RELATIVE ACCURACY TEST
                                     127

-------
(Table 10-4) and the fact that the sampling system used 1n this example

undersamples the finer particles.   The result Is that the sampling system

1s missing the portion of the coal stream with the highest sulfur

content.  This bias becomes more pronounced when there are more finer

coal particles 1n the coal  stream.  The opposite 1s true for larger coal

particles.

10.5  Bias Testing Using Emissions Monitors


     The previously described methods test for bias by automatic sampling

systems by comparing the results to a full cross-section cut from a

stopped belt.   An alternate reference sample could be obtained from an

emissions monitor (either CEM or Method 6B).  Dally SOa/MMBtu

calculations from both methods could be compared as paired observations

and the possible bias determined.   However, this procedure 1s not as

simple as the  direct comparison between two coal samples.  Possible

complications  are:

     1.  The unknown lag between the lot of coal which 1s sampled as 1t 1s
         bunkered and the measurement of pounds of SOa/MMBtu 1n the flue
         gas corresponding  to the  combustion of the same lot of coal.

     2.  The consistency of sulfur capture mechanisms may vary with
         changing coal characteristics, boiler operating conditions, and
         pollution control  device  efficiency.  Thus the relative
         difference between perfectly operating CSA and emissions monitor
         data  due to sulfur capture 1s not known.

     In our analysis below, we will make the following assumptions
regarding these Issues:

     1.  There 1s a twelve  hour delay between CSA measurements and
         emission monitor measurements on the same coal lot.  Thus the
         average of day  t and day  t+1 from the monitor represents the same
         coal  as that sampled on day t from the CSA procedures.

     2.  There 1s a constant depletion of s% and an allowed depletion
         credit of d% 1n the CSA measurement relative to the true coal
         property.  Note that we do not assume that s=d.

     3.  There 1s a negative bias  of B% 1n the CSA measurements relative
         to the true coal property.
                                   128

-------
     We use our model for the true coal Ib S02/MMBtu described above,
I.e. that the measurements may be split Into a "true process with error"
and an Independent white noise representing "measurement error."  From our
data analysis we further assume:
     4.  The measurement error for CSA, the measurement error for the
         emm1s1ons monitor, and the true coal process all have constant
         Relative Standard Deviations (RSD's), I.e.

                     standard deviation/measured mean = constant.

           Let

                    D1ff.  = (1-d) x Coal measurement on day t minus
                              Emissions measurement on day t.
Then
                    Expected Value (D1fft)  = [(s-d) - B(l-d)]w

where

                    v = true mean of the process.

     The relative accuracy of the CSA 1s measured by

                    D1ff./Emissions Data on day t.

If the bias 1s positive, I.e. the CSA procedures are producing systema-
tically high results, then the source 1s "cheating Itself."  If the bias
1s negative, I.e. the CSA procedures are producing systematically low
                                   129

-------
results, then 1t 1s possible that the source may be 1n violation of a
standard based on emission but Its CSA data would not so Indicate.  Thus,
a relative accuracy test should protect against large negative biases.  We
are therefore Interested 1n the probability that the relative accuracy of
the CSA methods be larger than some pre-asslgned (small) negative quantity:

                    A = Prob (-p < D1fft/Em1ss1onst).
Using the approximation
                  Emissions^ equals Its expected value,
A may be computed given values for the constants.
     A relative accuracy test would compare several days of CSA data to
the corresponding days for an emissions monitor.  If our test period Is
seven days and we require that the accuracy 1s above -p% for at least n of
these seven days, then the probability of passing such a test 1s
                           7
             Prob (pass) = I   -\"/• ;(7-.1+1)  AJ(1_A)7-J  .
                          J=n      Jt
These may also be tabulated.
     To select the optimal number for  n,  we need to make further
assumptions regarding the criteria for selection.  (5ur assumptions are:
     1.  At least 85% of unbiased samplers would pass the test.  The most
         difficult case to detect unbiased sampler 1s for the case s=o,
         I.e.  there 1s an allowance for sulfur depletion but there 1s no
         actual depletion.
     2.  We select that test which minimizes  the probability that a biased
         sampler with fixed bias  B would  pass the test.  The value n may
         depend on B.  The most difficult case to detect bias 1s for the
         case  s=smax, I.e. actual depletion 1s as large as  possible.
                                   130'

-------
     For all of our tables below we have assumed the following values:
         d=5%  (depletion allowance)
         o
-------
       TABLE 10-6.   PROBABILITIES OF CSA PASSING BIAS TEST
                      RELATIVE TO METHOD 68
 I.   B1as=0%,   s=0%,   d=5X

     Accuracy          A

        10%         .8002

        15%         .9540

        20%         .9942


II.   s=d=5%

     Accuracy-Bias    A

      -.05          .2092

       .00          .5000.

       .05          .7908

       .10          .9472
n=7


9604
n=6

.9619
.9993
n=5
.8523
.9970
1.0000
n=4
.9668
.9999
1.0000
n=7


.1933
.6839
n=6

.0625
.5514
.9509
n=5
.0057
.2266
.8357
.9956
" n=4
.0389
.5000
.9611
.9998
                              132

-------
       TABLE 10-7.   PROBABILITIES OF CSA PASSING BIAS TEST
                         RELATIVE TO CEM
 I.   B1as=0%,   s=0%,   d=5%

     Accuracy          A       n=7_        n=6_        n=5         n=4

        10%         .8194                          .8837        .9766

        15%         .9661                .9784       .9988       1.0000

        20%         .9969    .9786       .9998      1.0000       1.0000


II.   s*d=5%
Accuracy-Bias
-.05
.00
.05
.10
A
.1918
.5000
.8082
.9593
n=7


.2251
.7474
0=6

.0625
.5993
.9696
n=5
.0039
.2266
.8657
.9979
n=4
.0289
.5000
.9711
.9999
                              133

-------
the second test, an unbiased CSA would pass 97% of the time whereas a 15%
biased CSA would pass but 4% of the time.  Note that a 15% biased CSA
would pass the first test only 0.6% of the time while a 10% biased CSA
would pass the second test 50% of the time.
     We note that our selected tests would not change 1f the allowed
depletion were but 2-1/2% and the maximum theoretical depletion 1s either
2-1/2% or 5% (see Tables 10-8, 10-9).  The probabilities of passing such a
test would of course change.  A relative accuracy test compared to Method
68 with s    = 2-1/2% would have the following characteristics.
         max
           Test 1.  Unbiased CSA would pass 97% of the time while a
                    10% biased CSA would pass 23% of the time.
           Test 2.  Unbiased CSA would pass 99.7% of the time while a
                    15% biased CSA would pass 23% of the time.
If Smax = 5%, then the tests would be less sharp and would satisfy:
           Test 1.  Unbiased CSA would pass 97% of the time while a
                    10% biased CSA would pass 52% of the time.
           Test 2.  Unbiased CSA would pass 99.7% of the time while a
                    15% biased CSA would pass 18% of the time.
                                   134

-------
       TABLE 10-8.  PROBABILITIES OF CSA PASSING BIAS TEST

          RELATIVE TO METHOD 68 (ALTERNATIVE DEPLETION)
  I.   B1as=0y.,   s=0%.  d=2-l/2%



     Accuracy



         5%



        10%



        isy.



        20%





 II.   s=d=2-l/2%



     Accuracy-Bias



      -.05



       .00



       .05
             •


       .10





III.   s=5%,  d=2-1/2%



     Accuracy-Bias



      -.05



       .00



       .05



     ' .10
A
6602
8923
9805
9981
A
2092 .
5000
7908
9472
A
3386
6461
8782
9748
n=7


.8714
.9866
n=7

.0078
.1933
.6839
n=7

.0470
.4028
.8366
n=6

.8309
.9925
.9999
n=6
-
.0625
.5514
.9509
n=6

.2272
.7939
.9878
n=5

.9687
.9998
1.0000
n=5
.0057
.2266
.8357
.9956
n=5
.0484
.5233
.9567
.9995
n=4
.8166
.9964
1.0000
1.0000
n=4
.0389
.5000
.9611
.9998
n=4
.1814
.7936
.9943
1.0000
                              135

-------
     TABLE 10-9.   PROBABILITIES OF  CSA PASSING BIAS TEST
           RELATIVE TO CEH (ALTERNATIVE DEPLETION)
I.   B1as=0%,   s=0%,   d=2-l/2%
Accuracy
5%
10%
15%
20%
II. s=d=2-l/2%
Accuracy-Bias
-.05
.00
.05
.10
III. s=554, d=2-l/2%
Accuracy-Bias
-.05
.00
.05
.10
A
6722
9096
9871
9991

A
1918
5000
8082
9593

A
3275
6562
8945
9821
n=7


.9134
.9937

n=7

.0078
.2251
.7474

Q=7

.0524
.4583
.8814
n=6

.8736
.9967
1.0000

n=6
.0003
.0625
.5993
.9696

n-6
.0062
.2445
.8366
.9937
n=5

.9804
.9999
1.0000

n=5
.0039
.2266
.8657
•
.9979

n=5
.0420
.5465
.9704
.9998
n=4
.8352
.9981
1.0000
1 .0000

n=4
.0289
.5000
.9711
.9999

n=4
.1645
.8102
.9967
1.0000
                            136

-------
                                SECTION  11
                              DATA ANALYSIS
11.1   The Data
     We obtained  the  following CSA data  for analysis:
     1.  R&F Coal  Preparation  Plant feed  and  product  samples  collected at
         30 minute Increments  during two  Intensive efforts  for  22 and 40
         hours,  respectively,  during April-June, 1980.   Each  sample was
         split prior  to  laboratory analysis.
     2.  Republic  Steel  Coal  Preparation  Plant  feed and  product samples
         collected at 30 minute Increments during April-June, 1980.
     3.  Iowa Public  Service  coal  samples dally composites  during
         July-August, 1982.
     4.  Homer City as-fired  coal  samples at  15 minute Intervals for
         Units 1  and  2 and at  30 minute  Intervals for Unit  3  during
         November  1980-ApMl  1982  (Time  1).
     We obtained  the  following CEH data  for analysis:
     1.  Colorado  Public Service hourly  averages during  June-August, 1982.
     2.  Possum Point Generating Station  hourly averages  taken  with CSI
         and DuPont CEM's during March 1982.
     3.  Possum Point Generating Station  measurements taken with CSI and
         DuPont  CEM's at 15 minute Intervals  during February  1982.
     4.  Homer City dally averages at Units 1,  2, and 3  during  November
         1980-AprH 1982 (Time 1).
     5.  Homer City three hour averages  at Units 1 and 2  during December
         1981-February 1982 (Time  2).
     6.  Homer City dally averages at Unit 3  during December
         1981-February 1982 (Time  3).
     We obtained  the  followed  Method 68  data  for analysis:
                                   137

-------
     1.  Homer CHy three hour averages at Units 1  and 2 during December
         1981-February 1982 (Time 2).

     2.  Homer City dally averages at Units 1, 2, and 3 during December
         1981-February 1982 (Time 3).

     From these data, we obtained 53 useable data sets:  19 for CSA, 25
for CEM, and 9 for Method 6B.  A useable data set had to contain at least
45 (or so) consecutive data points.

11.2  A Typical Analysis

     One CSA data set we analyzed was the Republic  Steel clean coal data

set #2 consisting of 226 data points.  The data had the following
statistical characteristics:

                  average                   = 1.3795
                  variance                  = 0.0119
                  first order correlation   = 0.5746
                  second order correlation  = 0.4078.

The moment estimators for the auto-regressive and moving-average
parameters are (from (6-8) and (6-9)):

                  auto-regressive           =  .7097
                  moving-average            =  .2046.

More sophisticated estimators are obtainable via a  Box-Jenkins Time Series
routine.  It provided the estimators:

                  AR\X)                     =  .7998

                  MJl(X)                     =  .3679
with residual error variance, or white noise variance,

                  WNV(X)                    =  .00758.

     Versar Implemented a computer routine which computes the coal and
measurement error parameters based on these estimates.  The results are:

                  Estimates for the total process X
                     var(X)      =  .0115
                     AR(X)       =  .7998
                     MA(X)       =  .3679


                                   138

-------
                  Estimates for the coal process C
                  1  var(C)       =  .0080 = 69.55% of total variance
                    AR(C)        =  .7998
                    MA(C)        = -.00019
                    WNV(C)       =  .0029
                  Estimates for the measurement process ME
                    var (ME)     =  .0035 = 30.45% of total variance
                    RSO of error = 1/23
where
                  RSO of error   = standard deviation/estimated mean.
11.3  Goodness-of-F1t-Ana1ys1s
     To determine whether the fitted ARMA(l.l) model adequately fit the
data we considered three separate Indicators:  residual analysis, overflt,
and R2.
     The residuals are estimated by the difference between the forecast of
the next data point at each time and the actual observed data point:
                   resTdual = (Xt+1 - X) - ART/X) (Xt - X)
If the model fits, then these residuals should form an Independent
sequence of random variables.  If they do, then 9554 of the*1r correlations
should be 1n the Interval ±1.96^yTT-T"and
                      Q = (n) (sum of squared correlations)
should be small.  For our set above, the computer calculated the first
twenty correlations for the residuals.  We found that 2 of these 20 were
not 1n the Interval ±..12 and that Q=21.31.  Neither of these tests shows
that the ARMA(l.l) model 1s Inadequate (but, on the other hand, neither 1s
very compelling evidence either).
     The second method we used was "overflttlng" the model.  We fit the
data with ARMA(1,2) and ARMA(2,1) models to see whether either was
significantly better.  For the set analyzed above,  the white noise
variances were .0073 and .00745,  not significantly  better than .00758.
                                   139

-------
     The third Indicator of fit 1s the overall R2 defined by

         R2= 1-  tota1Uvar1ance"Ce = "'* of Var1ance explained by model."

In the above example R2 = .748.
     A summary of the model fits and goodness-of-f1t criteria 1s provided
1n Table 11-1 for our CSA data.  Summaries of the models for CEM and
Method 68 are provided by Tables 11-2 and 11-3.
11.4  Verification of the Coal Model
     Assuming that our ARNA(l.l) model fit the observed data, the next
Issue was whether the coal model described above adequately models the
real phenomena.  For this, we performed three analyses:  positive
variances, statlonarlty, and consistency.
     The split of the total variance Into two pieces, one due to the coal
and the other to the measuring process, 1s model dependent.  It Is possible
for other ARMA(l.l) models to distribute the total variance Into pieces
other than those obtained by our model.  The fact that we obtained two
positive values for the variances 1s, thus, a first "verification" of the
model.
     To study statlonarlty, the second Republic Steel Plant data were
split Into thirds:  the first group contains observations 1 to 88, the
second contains observations 89 to 176, and the third contains observations
177 to 264.  These three sequences are virtually Independent.  If, the coal
properties are stationary, then the mean and variance as well as the time
series parameters should be the same for these three series.

           Let y, and a* be the population mean and variance for the
1th series (1=1,2,3).  Estimates for these quantities are provided by the
sample means and variances:
                              mean                variance
                                                  .0098
                                                  .0113
                                                  .0117.
                                   140
Series 1
Series 2
Series 3
1.4096
1.3958
1.3348

-------
                                       TABLE 11-1.  SUMMARY OF RESULTS FOR CSA
A A AC (Residuals)

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

16.

17.

18.

19.

Data Set
R&F, A, ROM, Even Splits
R&F, A, ROM, Odd Splits
R&F, BC, ROM, Even Splits
R&F, BC, ROM, Odd Splits
R&F, A, Clean, Even Splits
R&F, A, Clean, Odd Splits
R&F, BC, Clean, Even Splits
R&F. BC, Clean, Odd Splits
Republic, Clean, Set 1
Republic, Clean, Set 2
Republic, ROM, Set 1
Republic, ROM, Set 2
Iowa P.S.
Homer City, Unit 1 (Time 1)
Homer City, Unit 2, Set 1
(Time 1)
Homer City, Unit 2, Set 2
(Time 1)
Homer City, Unit 3, Set 1
(Time 1)
Homer City, Unit 3, Set 2
(Time 1)
Homer City, Unit 3, Set 3
(Time 1)
Average
2.93
2.92
3.32
3.32
2.53
2.50
2.82
2.83
1.42
1.38
2.70
2.69
0.54
2.57

2.49

2.76

2.66

2.57

2.47
Var(C)
.0352
.0358
.0610
.0578
.0039
.0041
.0557
.0591
.0033
.0080
.0643
negative
negative
.0107

negative

.0326

negative

.0091

.0282
Var(ME)
.0109
.0096
.0668
.0594
.00945
.0087
.0314
.0316
.0050
.0035
.00375
split
split
.0095

split

.0294

split

.0147

.0440
RSD-KME) R2 out of Range1
28
29
13
14
26
27
16
16
20
23
44


26



16



21

12
.675
.705
.902
.911
.797
.805
.852
.869
.849
.748
.452


.619

-

.583



.794
*
.968
none
3
none
none
none
1
none
none
1
2
none
1
1
none

none

none

none

none

none
21
29
11
10
11
22
11
11
24
21
8
9
24
18

7

8

11

28

15
Q2
.5
.9
.6
.7
.7
.2
.2
.9
.7
.3
.2
.8
.2
.4

.3

.3

.9

.5

.0
Overfit
Improvement3
No
Yes
No
No
No
Yes
No
No
No
No
No
Yes
Yes
No

No

No

No

No

No

(Both)



(1,2)





(both)
(1,2)











1 Expected number out of range is 1,  i.e.  51 of 20 is  1.
2 Significance levels for 10V»26.0,  for 51=28.9,  for  1V=34.8.
3 Yes means white noise variance may be reduced by 10% by ARMA  (1,2) or ARMA  (2,1).
                                                       141

-------
                                    TABLE 11-2.  SUMMARY OF RESULTS FOR CEH


1.
2.
3.
4.
5.
6.
7.
8.

9.

10.

11.

12.

13.

14.

15.

16.
17.
18.
19.

20.

21.

22.

23.

24.

25.


Data Set
Colorado, Set 1
Colorado, Set 2
Colorado, Set 3
Colorado, Set 4
Colorado, Set 5
Colorado, Set 6
Colorado, Set 7
Homer City, Unit 1
(Time 1)
Homer City, Unit 2, Set 1
(Time 1)
Homer City, Unit 2, Set 2
(Time 1)
Homer City, Unit 3, Set 1
(Time 1)
Homer City, Unit 3, Set 2
(Time 1)
Possum Point, DuPont,
Set 1
Possum Point, DuPont,
Set 2
Possum Point, DuPont,
Set 3
Possum Point, CSI, Set 1
Possum Point, CSI, Set 2
Possum Point, CSI, Set 3
Homer City, Unit 1, Set 1
(Time 2)
Homer City, Unit 1, Set 2
(Time 2)
Homer City, Unit 1, Set 3
(Time 2)
Homer City, Unit 2, Set 1
(Time 2)
Homer City, Unit 2, Set 2
(Time 2)
Homer City, Unit 2, Set 3
(Time 2)
Homer City, Unit 3,
(Time 3)


Average
0.79
0.84
0.70
0.
0.
0.
0.

2.

2.

2.

2.

2.
1.

1.

1.

1.
1.
1.

2.

2.

2.

2.

2.

2.

2.
71
72
63
68

46

55

57

65

50
70

81

72

65
69
72

37

47

44

62

51

60

51

A
Var(C)
.0026
.0056
.00049
.0070
.0148
.0064
.00078

.0285

negative

negative

negative

.0265
.0375

.2177

.1113

.0243
.2095
.1032

.0198

negative

.0167

.0275

negative

negative

.0331

A
Var(ME)
.00026
.000027
.00016
.00041
.000016
.00015
.000014

.0165

split

split

split

.0194
.0035

.0027

.0008

.0030
.0091
.0008

.0005

split

.0002

.0016

split

split

.0057
AC (Residuals)
RSQ-HNE) R2 out Of Range 1 q2
5
162
55
35
180
51
182

19







18
29

35

60

30
18
60

106



173

66





33
.814
.868
.831
.914
.935
.902
.775

.700







- .878
.946

.855

.905

.883
.854
.905

.68



.61

.71





.24
none
none
1
none
none
none
none

1

none

none

none

1
1

none

none
•
2
1
none

none

none

none

none

1

2

1
6
26
12
15
18
19
20

13

10

16

9

18
19

8

16

41
12
9

8

9

13

9

9

15

12
.0
.2
.1
.3
.0
.7
.0

.7

.8

.8

.0

.5
.0

.2

.6

.6
.6
.2

.5

.5

.3

.3

.0

.4

.0
Overfit
Improvement3
No
No
No
No
No
No
No

No

Yes (1.2)

No

No

No
No

No

No

No
No
No

No

No

No

Yes (1,2)

No

No

No
1 Expected number out of range is 1,  i.e.  51 of 20 is  1.
2 Significance levels for 101=26.0,  for 51=28.9,  for 11=34.8.
3 Yes means white noise variance may  be reduced by 101 by ARMA (2,1) or ARHA  (1,2).

                                                       142

-------
                                       TABLE 11-3.  SUMMARY OF RESULTS FOR METHOD 6B
           Data Set
Average   Var(C)   Var(ME)
1.
2.
3.
4.
5.
6.
7.
8.
9.
Unit
Unit
Unit
Unit
Unit
Unit
Unit
Unit
Unit
1,
1,
1,
2,
2,
2,
1,
2,
3,
Set 1
Set 2
Set 3
Set 1
Set 2
Set 3
(Time
(Time
(Time
(Time
(Time
(Time
(Time
(Time
(Time
3)
3)
3)
2)
2)
2)
2)
2)
2)



2
2
2
2
2
2
2
2
2
.59
.60
.52
.76
.45
.59
.59
.76
.54
.
.
.
• .
.
.
.
.
0409
0305
0269
0247
0325
0249
0356
0236
Negative
.
.
,
.
,
.
.
.
0063
0029
0008
0025
0062
0014
0037
0372
Split
                                                                  33
                                                                  48
                                                                  89
                                                                  55
                                                                  31
                                                                  69
                                                                  43
                                                                  14
 R2

.81
.87
.66
.74
.55
.56
.25
.82
AC(Residuals)
out of Range1
  Overfit
Improvement^
1
1
none
none
none
none
1
none
none
24.4
9.7
19.5
13.1
11.2
14.0
18.7
8.8
13.8
Yes
No
No
No
Yes
No
No
No
No
(1,2)



(Both)




1 Expected number out of range is 1, i.e.  51 of 20 is 1.
2 Significance levels for 101=26.0,  for 51=28.9,  for lt=34.8.
3 Yes means white noise variance may be reduced by 101 by ARMA (2,1)  or  ARMA (1,2).
                                                      143

-------
An analysis of variance was performed to test the hypothesis of equal
means:
               Source       SS     d_f     HS     F
               Between     .97      2    .48    44
               Among      2.35    261    .01
Since the F-value 1s significant at the 99% level, the equality of the
means 1s rejected.  Further analysis showed that the mean of Series 3 was
significantly smaller than the other two while the other two were not
significantly different.  Stat1onar1ty of the mean apparently does not
hold for this long data set.   As a consequence,  the variance for the total
data set Includes variations  about local means and variation of the local
means.
     To test the hypothesis of equal variances Bartlett's test was used.
Here B=M/C 1s approximately *2(df=2)
                                   M = 1.802
                                   C = 1.005
and so
                                   B = 1.793,
This value 1s not significant and so equality of variances can be accepted.
Thus the mean but not the variance of the data was changing 1n time.
     The model parameters for the three series are:

                        A&(C)       VaV(C)          VaVME
          Series 1      .5149        .0075           .0018
          Series 2      .9366        .0065           .0037
          Series 3      .4794       Negative split.
          All data      .7998        .0080           .0035.

These values do not suggest that the time series model describing this
coal data set 1s time Independent.
                                   144

-------
     To test the consistency of the modeling procedure, the data were
split Into three groups:;  the first group contains the observations (Xi,
X4, XT,...), the second the observations (Xa, Xs, Xa,...), and
the third the observations (Xa.Xe.X*....).   The parameters
describing these series are:

                  Average   Sample variance     A^(C)     VaV(C)    virHE
    Series 1       1.37        .0124           .6621       .0082    .0035
    Series 2       1.36        .0104           .9214       .0041    .0063
    Series 3       1.38        .0131           .6047       .0082    .0050
    All data       1.38        .0115           .7998       .0080    .0035
Since the three series are highly correlated, standard tests are not
applicable.  Pa1rw1se testing whether the estimated means were equal
produced no significant differences.  (Here we used the appropriate
standard error of the mean of the differences from time series
considerations.)
11.5  Comparison of the Models for CSA and  CEH
     We obtained from Homer City measurements from their CSA and CEM
program for the same periods of time (listed above as Time 1).  Thus
determinations of the underlying coal process (when split off from the
measurement errors) was theoretically possible for two different media and
comparisons could be made.  Our results are presented 1n Table 11-4.
Their usefulness 1s limited by the following considerations:
     1.  CSA 1s performed as-fired rather than the typical as-bunkered.
     2.  About half the data sets had variance splits with negative
         measurement errors.
Using a pooled variance estimate and Independence (both Inappropriate) to
compare the averages, only Unit 3, Set  1  have estimates for the means
which are consistent.  Of the others, CSA 1s high three times and low once.
The estimates for the variance of the coal  are not consistent for the two
possible comparisons.  The estimates for the autocorrelation of the coal
are 1n reasonable agreement.
                                   145

-------
        TABLE 11-4.  COMPARISON OF CSA AND CEM. FOR HOMER CITY (TIME 1)
  Data Set
Unit 1, CSA
Unit I, CEM
Unit 2, Set 1, CSA
Unit 2, Set 1, CEM
Unit 2, Set 2, CSA
Unit 2. Set 2. CEM
Unit 3, Set 1, CSA
Unit 3. Set 1, CEH
Unit 3, Set 2, CSA
Unit 3, Set 2, CEM
iveraqe
2.57
2.47
2.49
2.55
2.76
2.57
2.66
2.65
2.57
2.50
Var(C)
.0107
.0285
**
**
.0326
**
**
Mr
.0091
.0265
ART.C) Sanple variance Nurber
.3332
.5886
.2911
.1232
.2515
.1137
.2444
.1776
.5232
.8252
.0284
.0470
.0268
.0225
.0623
.0632
.0132
.0175
.0238
.0455
63

90

82

58

156

                                     146

-------
11.6  Analysis of the R&F Data
     Laboratory samples for the R&F Coal Company Preparation Plant were
analyzed 1n duplicate thereby providing us with an opportunity to compare
two Independent determinations of the underlying coal  process.  Our
results are presented 1n Table 11-5.  Agreement between the split sample
streams for all four pairs 1s virtually perfect.
11.7  Measurement Error Analysis
     The R&F laboratory samples were analyzed 1n duplicate (see 11.6).
Thus, the difference 1n the paired measurements 1s a measure of the
laboratory analytic error, an error forming a part of  the total
measurement error.  An analysis of variance was performed on the four
paired data sets; results are listed 1n Table 11-6.  Averaging the results
1n Table 11-5 we have estimates:
                                                   Measurement Error
                                                         .0103
                                                         .0631
                                                         .0090
                                                         .0315
If we now Identify the between variance of ANOVA with  the laboratory
analytical error and the measurement error from above  with the sum of
sampling, preparation, and analytical errors, we may conclude the
following:
Data Set
A/ROM
BC/ROM
A/Clean
BC/Clean
Coal Variance
.0355
.0594
.0040
.0574

Data Set
A/ROM
BC/ROM
A/Clean
BC/Clean
Analytical
Error
.0045
.00265
.0044
.0043

(% of total)
(44%)
( 4%)
(49*)
(14%)
Sampling and
Preo Error
.0058
.06045
.0046
.0272

(% of total)
(56%)
(96%)
(51%)
(86%)
Thus we see that three data sets have equal  laboratory errors (of about
.0044) while the sampling plus preparation errors are extremely variable.
                                   147

-------
TABLE 11-5.  COMPARISON OF R4F SPLIT DATA SETS
Data Set
A/ROM/Even
A/ROH/Odd
BC/RON/Even
BC/ROH/Odd
A/Clean/Even
A/Clean/Odd
BC/Clean/Even
BC/Clean/Odd
Average
2.93
2.92
3.32
3.32
2.53
2.50
2.82
2.83
Var(C)
.0352
.0358
.0610
.0578
.0039
.0041
.0557
.0591
VAR\HE)
.0109
.0096
.0668
.0594
.0095
.0087
.0314
.0316
A^(C)
.6432
.6989
.8337
.8555
.3727
.4607
.8130
.8385
Hunter
44

62

41

82

                       H8

-------
          TABLE 11-6.  ANOVA FOR PAIRED R4F DATA
1. A/ROM Data
Source
Among
Between
2. BC/RON Data
Source
Among
Between

SS
4.34
0.20

SS
15.38
0.16

df
43
44

df
61
62

MS
. 10105
.00450

MS
.25208
.00265

Variance
.04828
.00450

Variance
.12472
.00265
3.  A/Clean Data

      Source

      Among
      Between

4.  BC/Clean Data

      Source

      Among
      Between
  SS

 1.20
 0.19
  SS

13.80
 0.35
df

42
43
df

79
80
   MS

.02847
.00436
   MS

.17468
.00432
Variance

 .01206
 .00436
Variance

 .08518
 .00432
                            149

-------
11.8  Conclusions
     The ARMA (1,1) model fits exceptionally well,  especially as 1t
accounts only for auto-correlation between measurements,  leaving out any
dependence on weather; season of year;  temperature, specifications, and
efficiency of the boiler; load; blending practices; spot  coal versus
contract coal supplies; etc.   It does about as well for data obtained via
CSA as 1t does for CEM and Method 6B generated data.   There 1s little
Indication of non-goodness-of-f1t (as measured by our analyses of
residuals) nor that an additional parameter would explain significantly
greater percentage of the variance.   All 1n all,  the  model  presented above
provides a very satisfactory, first  generation, universally applicable
model.
                                   150

-------
                                SECTION 12
                 CORRELATION OF CSA DATA TO S02 EMISSIONS

     The ultimate value of a CSA program will be based on the accuracy of
predicted S09 gaseous emissions for coal feedstock data.  The behavior
of the sulfur 1n coal must be considered from the sampling point to the
emission point.  If the parameters that affect the mass balance of sulfur
1n the system can be Identified, then an estimate can be made of the
sulfur loss.
     Sulfur loss 1n the pulverizers and 1n the boiler has already been
discussed (Section 5).  Parameters affecting sulfur loss at these sites
probably Include:  amount of pyrltlc sulfur 1n the coal, alkaline content
and composition of ash, fuel additives, boiler size,  boiler firing type,-
and boiler firing rate (percent excess air).
     A major factor that must be accounted for 1n the simulation of
sulfur loss 1s the S02 removal efficiency of pollution control
devices.  Systems-such as flue gas desulfurlzatlon units may trap a high
proportion of S0_.  Partlculate collection devices such as
electrostatic predpltators and mechanical dust collectors will not show
high SO- removal but may contribute by removing SO- or S03 adsorbed
to fly ash particles.
     Data sets which Include CSA data, emissions data, and boiler
operating parameters are virtually non-existent.  For this reason, no
causal analytic model could be developed.  In Heu of an explanatory
model, we developed the stochastic model described In Section 8.
                                   151

-------
                                REFERENCES
1.   Landry, B.A.  Fundamentals of Coal Sampling.  U.S. Bureau of Mines
     Bulletin 454.  U.S. Government Printing Office. Washington, O.C.
     1944.  pp 1-81.

2.   Meldner, A.C.  Notes on the Sampling and Analysis of Coal.  Techni-
     cal Paper No. 76, U.S. Bureau of Mines.  U.S. Government Printing
     Office.  Washington, O.C.  1914. pp 1-23.

3.   Pope, G.S.  Methods of Sampling Delivered Coal.  Technical Paper 116,
     U.S. Bureau of Mines.  U.S. Government Printing Office.   Washington,
     D.C.  1917.  pp 12-29.

4.   U.S. Steel Corporation.  Sampling and Analysis of Coal and Coke.
     Pittsburgh, PA.  1929. 334 pp.

5.   Snyder, N.H., revised by S.J. Aresco.  Coal Sampling. (Revision of
     Technical Paper 133).  U.S. Bureau of Mines.  U.S. Government
     Printing Office.  Washington, D.C. 1957.  pp 1-16.

6.   American Society for Testing and Materials.  1981 Annual Book of ASTM
     Standards.  Part 26.  Philadelphia, PA.  1981.  920 pp.

7.   Ornlng, A.A.  Coal Sampling Problems.  Presented at the  Fifty-Fourth
     Annual Meeting, American Society for Testing and Materials.
     Atlantic City, NJ.  June 18, 1951.  pp 29-35.

8.   Vlsman, J., S. J. Aresco, edited by J. W. Leonard.  Coal Preparation.
     The American Institute of Mining, Metallurgical, and Petroleum
     Engineers, Inc.  New York, 1979. pp 2-1 - 2-25.

9.   Bertholf, W.M.  The Design of Coal Sampling Procedures.   Presented at
     the Fifty-Fourth Annual Meeting, American Society for Testing and
     Materials.  Atlantic City, NJ.  June 18, 1951.  pp 46-57.

10.   Bertholf, W.M., W.L. Webb.  Tests of the Geary-Jennings Sampler at
     Cabin Creek.  Presented at the ASTM Symposium on Coal Sampling.
     Chicago, IL.  1954.  pp 83-111.

11.  Coryell, R.L., et.al.  Tests of Accuracy of a Mechanical Coal
     Sampler.  Presented at the ASTM Symposium on Coal Sampling.
     Chicago, IL. 1954.  pp 72-82.

12.  Gy, P. M.  Sampling or Gambling.  Coal Mining and Processing.
     September 1981.  pp 62-67.
                                   152

-------
13.  Gy, P. M.  Does Your Mechanical Sampler Do What It Is Suppose To?
     Coal Mining and Processing.  December, 1981.   pp 71-74.

14.  Gould, G., and J.  Vlsman, edited by R. A.  Myers.  Coal Handbook.
     Marcel Dekker, Inc.  New York.  1981.  pp  19-74.

15.  Fields, 0. K.  Plant Sampling.  Society of Mining Engineers of
     AIME.  Preprint No. 79-108.  1979.  pp 1-5.

16.  Huntlngton, F. R.   How to Select a Coal Sampling System.  Coal
     Mining and Processing.  March, 1982.  pp 42-45.

17.  Aresco, S. J., and A. A. Ornlng.  A Study  of  the Precision of Coal
     Sampling. Sample Preparation and Analysis.  Transactions, Society of
     Mining Eng1neers/AIME. Volume 232, 1965, pp.  258-264.

18.  International Organization for Standardization.  Solid Mineral Fuels.
     Technical Committee No. 27.  Geneva.  1981.

19.  Beaupre, E.V., Analysis of Ash Content and BTU of Coal.  American
     Laboratory.  Fa1rf1eld, CT.  November, 1980.   Vol. 12.  pp 89-93.

20.  Merrltt, Paul C. (editor), Coal Age Operating Handbook of Coal
     Preparation. Coal  Age Mining Informational -Services,  McGraw-Hill,
     Inc., New York, N.Y.

21.  Castald1n1, C. and M. Angwln.  Boiler Design  and Operating Variables
     Affecting Controlled Sulfur Emissions from Pulverized Coal-Fired
     Steam Generators.   Accurex Corp., Accotherm Division, U.S.
     Environmental Protection Agency, OAQPS, PB-281, 469.   Research
     Triangle Park, NC.  December, 1977.  80 pp.

22.  MITRE Corporation, Metrek Division.  Electric Utility Steam Generat-
     ing Units. Background Information for Proposed S02 Emission
     Standards.  U.S. Environmental Protection  Agency, OAQPS, EPA
     450/2-78-007a.  Research Triangle Park, NC.  1978.  488 pp.

24.  Homulya, James B.  and Scott Lambert.  Characterizations of Sulfate
     Emissions from Non-Utility Boilers Firing  Low -S Residual 011s 1n
     New York City.  Journal of the A1r Pollution  Control  Association.
     Pittsburgh, PA.  February, 1981.  Vol. 31, No. 2.  pp 139-143.

23.  Evers, Robert, V.E. Vandergrlff, and R. L. Zlelke.  Field Study to
     Obtain Trace Element Mass Balances at a Coal-Fired Utility Boiler.
     U.S. Environmental Protection Agency, Office  of Environmental
     Engineering and Technology, IERL, EPA-600/7-80-171.   Research
     Triangle Park, NC.  October, 1980.  143 pp.
                                   153

-------
25.  Crawford, A.R., E.H. Manny, W. Bartok.  Control of Utility Boiler and
     Gas Turbine Pollutant Emissions by Combustion Modification - Phase
     !_._  Report for June, 1944 - June, 1976.   U.S. Environmental
     Protection Agency, IERL, EPA-600/7-78-036a.  Research Triangle Park,
     NC.  1978.  150pp.

26.  Baker, Samuel S.  Electric Utility Coal  Sampling and Analysis
     Practices;  A Comparison to Proposed EPA Reference Method 19A
     Requirements Based on Utility Responses  to FERC Survey.  Kllkelly
     Environmental Associates, Inc.  Raleigh, NC.   Hay, 1982.  Report No.
     82-I155-03F.  26 pp.

27.  Cheng, B., K. CrumMne, A. Glelt, A. Jung, 0. Sargent, and
     B. Woodcock.  Variability and Correlation 1n  Coal:  Measurement and
     Analysis.  October, 1979.

28.  Cheng, B., A. Glelt, 0. Sarget, B. Woodcock.   Time Series Analysis
     of Coal Data from Preparation Plants.  JAPCA.  32 (1982), 1137-1141.

29.  Glelt, A., Technical Memorandom, Versar, 1983.

30.  ASTM, Proposed Test Method for Bias.  Comm. D-5 on Coal and Coke.
     October, 1982.

31.  Kaiser Engineers, Bias Test Results for  ROM Coal Sampler.  Pre-
     liminary Report.  March, 1982.
                                   154

-------
                     Table of Conversion Factors
   .Multiply
English Unit

 pound (Ib)

 ton (2000 Ib)


 inch (in.)

 foot (ft)

 British Thermal Unit
 (Btu)

 Btu/lb

 lb/106Btu

 gallon/minute  (gpm)
 453.59

   0.907


   0.0254

   3.048

1054.88


   2.326

 429.907

   0.06309
To Obtain
SI Unit

 gram

 megagram (Mg)
 metric ton

 meter (m)

 meter (m)

 Joule (J)


 J/g

 ng/J

 liter/second
 (1/s)
Tyler Screen Size Mesh Openings

        Mesh Size                 in.

           14                   0.0469
           24                   0.0234
           48                   0.0117
          100                   0.0059
          200                   0.0029
          270                   0.0021
          325                   0.0017
                   mm

                  1.18
                  0.60
                  0.30
                  0.15
                  0.075
                  0.053
                  0.045
                                 155

-------
APPENDIX A




DATA SETS

-------
  1.      DATA SET A. « 1 F COAL CO..  ROfl CSA DATA 30 KIN EVEN SPLITS
  3.               3.2636242
  3.               2.9491444
  4.               3.0006313
  5.               3.0714207
  6.               3.0636864
  7.               3.1259794
  8.               3.1471443
  9.               3.4042101
 10.               2.8887711
 11.               2.8985510
 12.               2.7754669
 13.               2.8903551
 14.               2.67I27SI
 IS.               2.56SI798
 1C.               2.4531479
 17.               2.7055359
 18.               2.8029585
 19.               2.9278SSS
 20.               2.9626493
 21.               3.0939808
 22.               2.9809713
 23.               3.2199974
 24.               3.0945129
 25.               2.5018396
 26.               2.8735628
 27.               2.9018784
 28.               2.6149960
 29.               2.71B6S08
 30.               2.5402193
 31.               2.7379808
 32.               2.6430311
 33.               2.8498774
 34.               2.8112440
 35.               2.9598303
 36.               3.3862429
.37.               3.1615553
 38.               3.1236668
 39.               3.1700S83
 40.               3.0926743
 41.               3.1203718
 42.               2.9746094
 43.               2.7716064
 44.               2.8477650
 45.               2.8762465

-------
  1.      DATA SET A. R 4 f COAL CO.. RON CSA DATA 30 HIM ODD SPLITS
  8.               3.337SS68
  3.               8.9883037
  4.               3.0538738
  5.               3.1840139
  6.               3.1161470
  7.               3.1782538
  8.               3.8409325
  9.               3.3308887
10.               8.9836578
11.               3.0811811
18.               8.7754669
13.               8.7646875
14.               8.5979468
IS.               8.6387720
16.               8.3794479
17.               8.9152679
IB.               8.8029585
19.               8.7877568
80.               3.1813617
81.               3.0788607
88.               8.9069757
83.               3.0883068
24.               8.9791250
25.               8.5438871
26.               8.7795191
87.               8.7766171
88.               8.5834964
89.               8.6838136
30.               8.6354780
31.               8.6106329
32.               2.6854887
33.               8.6591797
34.               8.8118440
35.               3.0549679
36.               3.8698418
37.               3.0877857
38.               3.1663113
39.               3.0958433
40.               3.0926743
41.               3.1203718
48.               8.9108667
43.               8.8456564
44.               8.8865924
45.               3.0036077

-------
CJ
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
IS.
16.
17.
18.
19.
20.
21.
22.
23.
24.
as.
26.
87.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
St.
51.
52.
S3.
DATA SETS t-C,
3.5984430
3.6167717
3.4803028
3. 451 1957
3.6S9184S
3.3788099
3.2048SS0
3.1134167
3.1S93704
8.4768124
3.4567366
2.9732304
3.1217041
3.8105761
3.2181867
3.S278006
3.7807579
3.9327621
3.2757339
3.6S64798
3.9331779
3.6228809
4.0679760
4.0063985
3.S063S68
3.5898873
3.4511776
3.4767857
3.1717396
3.7640095
3.8189518
3.8679739
3.8813764
3.0568601
3.4713373
3.4988026
3.5778360
3.S44S18S
3.6887846
4.241S371
6.0S99S66
3.2423201
2.9302073
3.3033342
2.7899990
3.0067987
3.2017069
3.018SS09
3.6507931
3.1292801
3.0417083
8.7597914
                                                                               64.
                                    ft I F COAL CO..ROM CSA DATA 3» HIN EVEN SPtHS
                                                                               56.
                                                                               57.
                                                                               SB.
                                                                               59.
                                                                               60.
                                                                               61.
                                                                               62.
                                                                               63.
                                                                               64.
3.8637482
2.8709631
 .1486454
 .8777733
 .•691071
 .6014891
 .3228006
3.0453625
3.1944151
2.S0S0411
3.4937725

-------
DATA SETS B-C, R I F COAL CO.,RON CSA DATA 36 HIN ODD SPLIII
1.
2.
3.
4.
S.
6.
. 7.
8.
9.
19.
it.
ia.
13.
14.
IS.
16.
17.
18.
19.
20.
81.
28.
83.
84.
85.
86.
87.
88.
89.
3*.
31.
38.
33.
34.
35.
36.
37.
38.
39.
40.
41.
48.
43.
44.
45.
46.
47.
48.
49.
50.
St.
sa.
S3.
54.
DATA SETS 1
3.5110944
3.6061020
3.4869238
3.4088497
3.5315388
3.3361750
3.3219757
3.1663656
3.8132835
8.5189714
3.4847303
3. 0049725
3.1111574
3.1790991
3.3817444
3.6225767
3.7490759
3.8799019
3.3393393
3.6564798
3.8591661
3.6440668
4.0148687
4.0S966SS
3.5487889
3.5088169
3.5248318
3.5085859
3.1399164
3.6688519
3.1833387
3.3839340
3.1474488
3.9985613
3.3970870
3.4398813
3.6200583
3.4914560
3.6993847
4.0848303
6.0599566
3.3063135
2.8449650
3.8398090
8.9187679
3.0599238
3.1696901
2.9451838
3.4814806
3.0133801
3.0417083
2.781*202
3.2319593
II.
56.
57.
58.
59.
60.
61.
62.
63.
64.
2.902S13S
3.0772438
2.9414883
3.1749382
2.5490398
3.3024149
3.1511040
3.3091900
2.4519691
3.4937725

-------
 1.      DATA SET A. R I F COAL CO..  CLEAN COAL CSA DATA 30 tllN EVEN SPLITS
 8.               8.9494543
 3.               8.7387814
 4.               8.4894886
 5.               8.4883871
 6.               8.7484746
 7.               8.5550658
 8.               3.5886851
 9.               8.4683060
10.               8.4363585
11.               8.4097643
18.               8.4958944
13.               8.8855030
14.               8.3488188
IS.               2.5635567
16.               2.4769764
17.               8.5498488
18.               8.6874300
19.               8.5851198
80.               8.5568801
81.               8.5874853
88.               8.5567875
83.               8.4567690
84.               2.6681419
85.               8.3588897
86.               8.6174698
87.               8.4418354
88.               8.3695407
89.               8.3488518
30.               8.6343403
31.               8.5571146
38.               8.5866499
33.               8.6161861
34.               8.4949350
35.               8.8107634
36.               8.4770737
37.               2.5751978
38.               8.6701841
39.               8.5711851
40.               8.6575818
41.               8.4909897
48.               8.3613685
43.               8.4876661
44.               8.5366868

-------
              t.      DATA SET A. R I F COAL CO.. CIEAN COAL CSA DATA 36 HIN ODD SPLITS
              8.               3.7S36442
              3.               2.6336088
              4.               8.4186378
              S.               8.4336893
              6.               3.5863981
              7.               8.3788538
              8.               8.4476976
              9.               2.S16S4IS
             10.               8.3271055
             it.               3.3676789
             IB.               8.3973788
             13.               8.341S8SI
             14.               8.4228546
             IS.               2.6913622
             16.               8.4981476
             17.               8.4689379
             18.               8.6799784
             19.               8.4806389
             86.               8.5878391
             81.               8.6465687
             88.               8.5779181
             83.               8.4989458
X»           84.               8.5657336
~           85.               8.3159885
             86.               8.4913873
             87.               2.3951155
             88.               8.3886933
             89.               8.3864568
             36.               8.44S4831
             31.               8.3894358
             38.               8.4981751
             33.               8.5949154
             34.               8.5655971
             35.               8.8543466                   *
             36.               8.5198366
             37.               8.5546895
             38.               8.5963658
             39.               8.5186616
             46.               8.5616866
             41.               8.4969897
             48.               2.4575272
             43.               8.5401487
             44.               8.5694894

-------
 t.       DATA SETS B AND C, H I
 a.                2.6281767
 3.                3.8796368
 4.                3.0981989
 S.                a.61*5251
 6.                8.6932831
 7.                £.5126152
 8.                8.7850752
 9.                2.7489243
1*.                2.3894739
II.                2.8799658
12.                3.1952906
13.                3.1600075
14.                3.2S512S0
IS.                2.8535185
16.                3.2397404
17.                2.7391434
18.                2.833S800
19.                3.0373335
20.                2.4832354
21.                2.7137461
22.                2.6840010
23.                2.4405565
24.                2.4433718
25.                a.7001562
26.                2.4SS0772
27.                2.5952101
28.                2.8835917
29.                2.7282267
30.                2.8376961
31.                2.6542168
32.                2.5437851
33.                2.8164616
34.                3.2341919
35.                3.0092344
36.                3.0912848
37.                2.9959421
38.                3.1336568
39.                3.2493715
40.                3.0096769
41.                3.3272133
42.                2.8994570
43.                3.0281029
44.                3.0381584
45.                3.2040644
46.                3.0S9833S
47.                2.4234819
48.                2.5597973
49.                3.2198199
50.                2.6332293
St.                2.7312931
S2.                2.7684565
S3.                3.3732424
                                 F COAL CO.. CLEAN COAL CSA DATA
 54.
EBgN SPLITS
 56.
 57.
 58.
 59.
 60.
 61.
 62.
 63.
 64.
 65.
 66.
 67.
 68.
 69.
 70.
 71.
 72.
 73.
 74.
 75.
 76.
 77.
 78.
 79.
 80.
 81.
2.8466377
3.1897087
3.0694838
3.580BS92
3.3193712
3.5960684
2.8778477
2.8266268
2.77S40S9
2.6920614
2.4325752
2.5489779
2.5991793
2.7570238
2.5164452
2.5230761
2.5525694
2.6568031
3.5119858
2.4277935
2.2336407
  4046431
  6014891
  6631422
  6573572
  6302099
  7049522
                    2.6461658

-------
I
oo
 1.       DATA SETS I AND C, R
 8.                2.7851177
 3.                8.7508155
 4.                3.6874777
 S.                8.6842098
 6.                8.7355299
 7.                2.6848243
 8.                8.7640568
 9.                8.7489243
I*.                8.4848100
It.                2.9226085
18.                3.4166063
13.                3.2234241
14.                3.3291054
IS.                2.9169308
16.                3.3261328
17.                8.7391434
18.                8.7912874
19.                3.4795183
20.                8.5565796
21.                8.7432681
22.                8.6210937
23.                8.5767254
24.                2.3175325
25.                8.6268969
26.                2.6413365
27.                2.5952101
88.                8.8623104
29.                8.7072441
30.                2.7434S49
31.                2.8280720
32.                2.6167641
33.                2.9203901
34.                3.8133948
35.                2.8743388
36.                3.0088979
37.                3.1103706
38.                3.1308041
39.                3.1512375
40.                3.1479311
41.                3.4304714
42.                2.9933243
43.                2.9451448
44.                3.1314848
45.                3.8440644
46.                2.9969597
47.                2.4966049
48.                2.4758701
49.                3.1769449
50.                8.6854744
51.                8.6158448
58.                8.7160834
S3.                3.3484486
                                            i F COAL CO.. CLEAN COAL CSA
      54.
DATA 000.SPLITS
      56.
      57.
      58.
      59.
      60.
      61.
      62.
      63.
      64.
      65.
      66.
      67.
      68.
      69.
      70.
      71.
      72.
      73.
      74.
      75.
      76.
      77.
      78.
      79.
      80.
      81.
8.7731094
3.2001677
3.1746035
3.6442375
3.3193718
3.7815137
2.7833204
2.8540850
2.9248495
2.7C65462
8.6335868
8.4225826
2.5676088
8.6683163
8.5373887
2.5230761
2.7437296
8.4901018
8.4390840
8.5010462
2.43288S2
2.3528194
2.5490398
3.7139654
8.6365147
2.5574379
8.5384928
8.6461658

-------
 I.      REPUBLIC STEEL CLEAN CSA DATA 30 IIIN  1ST SET
 2.                 .4340332
 3.                 .4249333
 4.                 .5850778
 S.                 .6992569
 6.                 .4343386
 7.                 .4138888
 8.                 .5550747
 9.                 .5443763
10.                 .4987135
11.                 .4539757
12.                 .4799604
13.                 .4359472
14.                 .5041599
IS.                 .4934921
16.                 .3051443
17.                 .4130087
18.                 .4480124
19.                 .4262381
20.                 .3574648
21.                 .3353930
22.                 .3633928
23.                 .4063339
24.                 .4008026
25.                 .3899441
26.                 .3525114
27.                 .4061832
28.                 .4299631
29.                 .4732952
30.                 .4188356
31.                 .4858665
32.                 .5188789
33.                 .4440050
34.                 .3545723
35.                 .4629393
36.                 .2574596
37.                 .4599314
38.                 .3311682
39.                 .4826832
40.                 .3551903
41.                 .5191259
48.                 .4674368
43.                 .3924589
44.                 .4002838
45.                 .4002838
46.                 .4236403
47.                 .4236403
48.                 .5288467
49.                 .4507303
S0.                 .5074816
SI.                 .5294895
52.                 .4004364
S3.                 .4770241
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
80.
81.
82.
83.
84.
85.
86.
87.
88.
89.
90.
91.
.3442039
.3772583
.4347820
.5108690
.5333355
.4801397
.5614967
.5614967
.4463357
.5124741
.3344898
.3561888
.4585629
.3149166
.5381279
.4298086
.4323090
.4322090
.3613710
.2396231
.2292433
.2184601
.2998457
.2667980
.3937584
.3046303
.3966475
.5148249
.1976042
.1976042
.5109291
.5645084
.3262596
.3050385
.5366545
.5366545
.3867817
.3109417

-------
 1.      REPUBLIC STEEL CLEAN CSA DATA 30 HIN 2ND SET
 8.                 .6143635
 3.                 .445451?
 4.                 .5036783
 5.                 .3330488
 6.                 .3674650
 7.                 .3018255
 8.                 .3944006
 9.                 .4694834
10.                 .3457670
II.                 .4305391
13.                 .3674450
13.                 .3123064
14.                 .3618460
IS.                 .3299284
16.                 .3999357
17.                 .4105415
18.                 .5112810
19.                 .4048519
20.                 .3330507
21.                 .3224707
22.                 .3974161
23.                 .3974161
24.                 .4066689
25.                 .3749332
26.                 .3957481
27.                 .3742743
28.                 .2741547
29.                 .2741547
30.                 .5278673
31.                 .4740686
32.                 .4726133
33.                 .3984528
34.                 .5002699
35.                 .4355097
36.                 .3632965
37.                 .3313446
38.                 .2826033
39.                 .3688297
40.                 .2346973
41.                 .2870140
48.                 .4661808
43.                 .3912668
44.                 .3212566
45.                 .4704313
46.                 .4055796
47.                 .3626614
48.                 .3839712
49.                 .4054288
50.                 .3908062
Si.                 .4438896
S3.                 .3117523
S3.              .   .3434887
 54.
 55.
 56.
 57.
 58.
 59.
 60.
 61.
 62.
 63.
 64.
 65.
 66.
 67.
 68.
 69.
 70.
 71.
 72.
 73.
 74.
 75.
 76.
 77.
 78.
 79.
 80.
 81.
 82.
 83.
 84.
 85.
 86.
 87.
 88.
 89.
 90.
 91.
 92.
 93.
 94.
 95.
 96.
 97.
 98.
 99.
100.
101.
102.
103.
104.
105.
.3269253
.3269253
.4048262
.4798927
.4850969
.4331703
.2752171
.2438583
.5537796
.4443588
.3221912
.3539238
.5136223
.5388498
.3614120
.3933201
.3608332
.3183069
.4302874
.4739599
.3928757
.5098343
.1308727
.3285017
.3912849
. 3589296
.5302792
.4217491
.5005808
.S2I71S2
.6596613
.7243233
.5075369
.5874825
.6159372
.5938005
.4967766
.4749260
.4187489
.5169706
.6632938
.6632938
.4486380
.5035114
.6334133
.5785999
.5658922
.6344719
.4789658
.5008755
.5421629
.5531006

-------
196.
107.
tee.
109.
110.
in.
118.
113.
114.
115.
116.
117.
118.
119.
180.
181.
188.
183.
184.
IBS.
186.
187.
188.
189.
130.
131.
138.
133.
134.
135.
136.
137.
138.
139.
140.
141.
148.
143.
144.
145.
146.
147.
148.
149.
150.
151.
158.
153.
154.
155.
156.
157.
158.
.4888886
.5161488
.3938408
.3788315
.3399889
.3399889
.5710388
.4116564
.4687605
.4009705
.3453903
.3895016
.4669657
.8849331
.3751554
.4098503
.3788576
.3898010
.4098358
.3568893
.5313063
.4178768
.4319544
.4319544
.3663908
.3884896
.3987869
.4858868
.3689150
.4368715
.8619895
.8619895
.3041639
.1989899
.2915716
.2869936
.3595161
.3595161
.8445011
.8445011
.3836548
.3770875
.8893677
.8576609
.8808785
.8375965
.8819954
.8648786
.1900959
.8644768
.3606888
.3713188
.3139763
159.
160.
161.
168.
163.
164.
165.
166.
167.
168.
169.
170.
171.
178.
173.
174.
175.
176.
177.
178.
179.
180.
181.
188.
183.
184.
IBS.
186.
187.
188.
189.
190.
191.
198.
193.
194.
195.
196.
197.
198.
199.
800.
201.
208.
803.
204.
305.
206.
207.
208.
809.
210.
811.
818.
.3139763
.4885416
.4885416
.4006786
.4431829
.3178701
.8335268
.3566923
.3994286
.3442869
.3549557
.4558784
.4871454
.4171543
.3958445
.3899117
.3793831
.4005604
.4436541
.2538519
.3888388
.2670250
.2775841
.2561731
.1708805
.3913965
.3388893
.8996769
.8781944
.3016996
.8498113
.3668395
.3111496
.2950388
.3785896
.3692814
.4578533
.5208311
.6386414
.3986006
.504S557
.4833918
.4833918
.5604095
.3597860
.3931783
.5101557
.8651548
.3600416
.2451200
.8668830
.4359188
.4680668
.4030199

-------
IV)
813.
814.
815.
816.
817.
818.
819.
880.
881.
888.
883.
884.
885.
886.
887.
888.
889.
830.
831.
832.
833.
834.
835.
836.
837.
838.
839.
840.
841.
848.
843.
844.
845.
846.
847.
848.
849.
850.
851.
858.
853.
854.
855.
856.
857.
858.
859.
864.
861.
862.
863.
864.
865.
.4430199
.3589798
.3798866
.3044713
.8794968
.1489176
.8474065
.4938148
.1148514
.3181499
.8945998
.2818437
.2968170
.8881856
.3032045
.3176818
.3069496
.1816368
.1816368
.2539511
.2855635
.3196478
.3884873
.1668834
. 1983604
.4064316
.4388305
.3069773
.3393818
.2931494
.8931490
.5248311
.4779908
.4778318
.4778318
.2664795
.2457170
.3465471
.3788633
.1188517
.1821823
.5153227
.5153227
.4031601
.3589649
.2765055
.2026587
.3235134
.8914925
.3929377
.3929377
.4411641
.3584513
                                                                              266.
                                                                              267.
                                                                1.3331894
                                                                1.2941831

-------
 I.      REPUILIC STEEL BOH CSA DATA 3* HIN isr SET
 8.               2.1851S6B
 3.               3.1534888
 4.               3.36S3S63
 S.               3.8738714
 6.               8.6835117
 7.               8.7643953
 8.               2.3379049
 9.               3.3171335
16.               3.6875139
it.               8.9316588
18.               8.3863683
13.               8.3993326
14.               8.3663781
15.               8.2833414
16.               8.S664711
17.               2.5562648
18.               8.1552162
19.               2.2173862
26.               8.3168639
21.               2.3587376
22.               2.8982477
23.               2.7731438
84.               8.8827868
85.               8.8465485
26.               8.4952618
27.               2.6216646
88.               2.3246669
89.               2.4397898
36.               2.7559891
31.               2.7841879
32.               2.8867273
33.               2.9285631
34.               2.6887932
35.               2.6887932
36.               8.9686357
37.               3.1133842
38.               2.5433769
39.               2.BS59163
46.               2.4953263
41.               2.5369691
48.               2.9836576
43.               3.6638452
44.               8.5623556
45.               8.5861158
46.               2.6447836
47.               2.6447836
48.               2.6259832
49.               2.4239836
5*.               3.3666841
St.               3.3813678
58.               3.1187649
S3.               3.1187649
54.
55.
56.
57.
58.
59.
66.
61.
68.
63.
2.7468983
8.7462983
2.1964755
8.8328633
2.9969597
2.9969S97
3.3487689
3.1646833
8.5532351
2.6255941

-------
 1.      REPUBLIC STEEL RON CSA DATA 30 HIM 7«D SET
 8.               3.0347789
 3.               8.8977871
 4.               2.6588984
 5.               8.8177958
 6.               3.0149155
 7.               8.9514437
 8.               3.esiea44                   •
 9.               3.0299835
10.               a.8098818
li.               3.8735618
13.               8.8073158
13.               a.9987330
14.               a.8434525
IS.               3.8434525
16.               2.8318758
17.               a.8318758
18.               2.8451881
19.               a.9393301
80.               8.4778654
31.               2.3727398
22.               2.9696590
23.               2.3232107
24.               3.0426588
85.               3.9112915
26.               2.4461937
27.               2.5731779
a8.               8.4239883
29.               3.4448853
30.               3.7876560
31.               2.7870560
32.               3.1581125
33.               3.1685019
34.               2.5463686
35.               2.6511574
36.               2.9839725
37.               2.8785315
38.               3.2I600S3
39.               3.0169697
40.               2.2834396
41.               2.440SS6S
48.               8.9160585
43.               2.9160585
44.               2.8895655
45.               a.9315958
46.               8.3043385
47.               a.3043385
48.               8.9319696
49.               3.7337887
56.               3.1944876
St.               3.1109705
52.               8.6795956
S3.               2.5544777
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
8.9175472
8.8858337
3.1820722
3.1949205
8.9730301
3.0261202
2.6216831
2.7574682
2.8993139
2.7938843
2.5238237
8.6180744
2.9177990
3.0537539

-------
3*
I
 I.       IOUA PUBLIC SERUICE CSA DATA
 8.                .3780000
 3.                .5530000
 4.                .5450000
 s.                .4080000
 6.                .4310000
 7.      .        0.5080000
 8.               6.4978868
 9.               8.4719688
18.               6.4288668
It.               8.3978668
18.               8.4718668
13.               8.4418688
14.               e.4394800
IS.               0.4830000
1$.               0.5370000
17.               0.4770000
18.               0.4650000
19.                .4840000
80.                .5090000
21.                .4290000
88.                .3880000
83.                .3750800
84.                .3210000
25.                .3810000
86.                .3780000
87.                .3300000
88.                .3360800
29.                .3590000
30.                .3980000
31.                .4970000
38.                .5840000
33.                .5480000
34.                .5870000
35.                .5050000
36.                .5390000
37.                .6810000
38.                .5930000
39.                .5560000
40.                .5380000
41.                .5380000
48.                .6850000
43.                .6730000
44.                .7180000
45.                .7110000
46.                .5380000
47.                .6300000
48.                .6780000
49.                .7890000
50.                .6700000
51.                .6560000
68.                .7450000
S3.                .6340000
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
78.
73.
74.
75.
76.
77.
78.
79.
80.
81.
88.
83.
84.
85.
86.
87.
88.
89.
90.
91.
92.
93.
 .6480000
 .8600000
 .9280000
 .6250000
 .5830008
 .6190000
 .5688080
 .6340000
 .6390088
 .6398880
 .5110000
0.4750860
0.4800000
0.6010000
0.6330000
0.7830000
0.7120000
0.6200000
0.5808000
0.6580000
0.7300000
0.6970000
0.5790000
0.5400000
0.6230000
0.5600000
0.5360000
0.4890800
0.4610000
0.4280000
0.4760000
0.4710000
0.4630000
0.4940080
0.4658000
0.4120000
0.4700000
0.4230000
0.4400000
0.4040080

-------
                                                                  54.
 I.
 3.
 3.
 4.
 S.
 6.
 7.
 8.
 9.
10.
II.
18.
13.
14.
IS.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
SI.
52.
S3.
' PLANT UNIT
3.8399998
2.8599997
2.3999996
2.7700005
2.4799995
2.8199997
2.5799999
2.6499996
2.4899998
2.5400000
2.5699997
2.4399996
2.7200003
2.5699997
2.8000002
2.7299995
2.5000000
2.4799995
2.3900003
2.3699999
2.3800001
2.5200005
2.7600002
2.6400003
2.6099997
2.5500002
2.4300003
2.6199999
2.6599998
2.7700005
2.8000002
2.7200003
2.S699997
2.5900002
2.7100000
2.1800003
2.3699999
2.6300001
2.5000000
2.5900002
2.4600000
2.4200001
2.4099998
2.5100002
2.5799999
2.6499996
2.5500002
2.4300003
2.4899998
2.6199999
2.790000*
2.6700401
1 CSA AND CEH DATA DAILY OUERLAPPINGSS
56
57
58
59
60
61
62
63
64
•>





















'




















2.5400000
2.5600004
2.3999996
2.5100002
2.3800881
S.2988000
2.6400003
2.6400003
2.4399996
2.2600002
2.4200001

-------
 I.
 8.
 3.
 4.
 5.
 6.
 7.
 8.
 9.

II.
18.
13.
14.
IS.

17.
18.
19.
80.
81.
88.
83.
84.
85.
86.
87.
88.
89.
30.
31.
38.
33.
34.
35.
36.
37.
38.
39.
40.
41.
48.
43.
44.
45.
46.
47.
48.
49.
SO.
SI.
58.
S3.
HOHER CITV PLANT UNIT 8 CSA AND CEH DATA
          8.5800005
          8.4099998
          8.3699999
          8.0900008
          3.1599998
          8.1999998
          8.8900000
          8.3800001
          8.4099998
          3.0699997
          8.8000008
          8.6099997
          8.4300003
          8.3699999
          8.4899998
          8.6400003
          8.8199997
          8.6499996
          8.5400000
          8.6199999
          8.7399998
          8.8899999
          8.6999998
          8.5799999
          8.4690000
          8.3999996
          8.5100008
          8.5000000
          8.6199999
          8.4700003
          8.4799995
          8.4600000
          8.8700005
          8.4499998
          8.3899999
          8.4099998
          8.3800001
          8.6499996
          8.5000000
          8.4799995
          8.5100008
          8.6700001
          8.5600004
          8.39000*3
          8.4799995
          8.3599997
          8.3SOO**4
          8.S900M8
          8.S9MM8
          8.4499998
          8.5699997
                 54.
DAILV OUERLAPPINCSS.
                 56.
                 57.
                 58.
                 59.
                 60.
                 61.
                 68.
                 63.
                 64.
                 65.
                 66.
                 67.
                 68.
                 69.
                 70.
                 71.
                 78.
                 73.
                 74.
                 75.
                 76.
                 77.
                 78.
                 79.
                 80.
                 81.
                 88.
                 83.
                 84.
                 85.
                 86.
                 87.
                 88.
                 89.
                 90.
                 91.
8.6199999
8.I300001
8.3900003
8.5600004
8.6000004
8.6999998
8.6099997
8.4600000
8.6400003
8.6199999
8.5000000
8.5799999
8.5000000
8.3599997
8.4399996
8.4499998
8.3599997
8.4800001
8.4800001
8.3000008
8.4700003
8.4600000
8.6999998
8.6000004
8.6199999
8.5000000
8.4300003
8.4499998
8.3190004
8.4499998
8.6899996
8.4899998
8.1400003
8.0600004
8.4499998
8.5399997
8.3599997
8.5800005

-------
I
_4
00
 1.
 a.
 3.
 4.
 5.
 6.
 7.
 8.
 9.
10.
It.
13.
13.
14.
IS.
16.
17.
18.
19.
ao.
zi.
aa.
83.
84.
as.
36.
87.
88.
89.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
43.
43.
44.
4S.
46.
47.
48.
49.
SO.
SI.
58.
S3.
HOHER CITV PLANT UNIT 3 CSA AND CCH DATA DAILV
          8.6304801
          a.2seaee«
          8.8899999
          8.8989693
          8.520000s
          8.8000002
          2.8400002
          2.8406068
          8.8406068
          8.770000S
          8.9706003
          2.7799997
          8.9399996
          3.1699997
          3.8666008
          3.1366001
          8.7399998
          2.7900000
          8.5900002
          2.7166600
          3.8706665
          8.7266663
          8.3866661
          2.6066604
          8.7706665
          3.1699997
          8.7606008
          8.8506004
          8.8500004
          3.6E60004
          2.8466002
          8.8800001
          8.8406068
          8.8506604
          8.8599997
          8.7966660
          8.9600600
          3.0806665
          8.8506000
          3.6799999
          2.2700065
          2.4306003
          8.4600600
          8.6066004
          8.5600004
          8.8466008
          8.8166004
          3.1699997
          2.8699999
          2.7799997
          8.6999998
          2.6099997
           54.
2ND OUCRLAPBIMC
           56.
           57.
           58.
           59.
           66.
           61.
           62.
           63.
           64.
           65.
           66.
           67.
           68.
           69.
           76.
           71.
           78.
           73.
           74.
           75.
           76.
           77.
           78.
           79.
           86.
           81.
           88.
           83.
8.5100008
8.6300661
2.6999998
2.6366661
8.6199999
8.5799999
8.5699997
8.4306003
8.5600664
8.9800001
2.8500664
2.6199999
8.5966662
8.8199997
2.8860602
2.4600000
2.8800001
8.7800003
2.9099998
2.9499998
8.5100002
8.7500000
8.7500000
8.5699997
8.6700001
3.3699999
8.6899996
8.8399998
8.5800005
8.9399996

-------
                                                                               54.                8.6899996
              1.     HOMER CITY PLANT UNIT 3 CSA CEH DATA DAILV 1ST OUERLAPPINGSS.                8.9300003
              8.               8.6300001                                       56.                8.7799997
              3.               a.6988804                                  .     57.                8.6400003
              4                8 7500000                                       5B-                8.5900008
              S.'               8.7700005                                       59.                8.6099997
              6.               8.6599998                                    '
              7.               8.6099997
              8.               8.7100000
              9.               8.8199997
             10.               8.6599998
             11.               8.4099998
             18.               8.4099998
             13.               8.7399998
             14.               8.6199999
             IS.               8.7700005
             16.               8.5799999
             17.               8.6700001
             18.               8.5400000
             19.               3.5388885
             80.               8.6400003
             81.               8.5799999
             88.               8.5390005
^           83.               8.7399998
I             84.               8.7600008
^           85.               8.6499996

             8?!               8.6499996
             88.               8.5900008
             89.               8.6300001
             30.               8.6899996
             31.               8.8890001
             38.
             33.
             34.               8.7700005
             35.               8.8900003
             36.               8.8500004
             37.               8.8100004
             38.               8.6499996
             39.               8.7899995
             40.               8.6199999
             41.               8.5699997

             43.               8.6499996
             44.               8.7600008
             45.               8.6199999
             46.
             47.
             48.               8.5799999
             49.               3.5799999
                               M9 C^A^MkAA
                •               E»9^VWW
             St.               8.5699997
             58.               8.4799995
             S3.               8.48M001

-------
I
f\>
o
 t.
 a.
 3.
 4.
 5.
 6.
 7.
 8.
 8.
18.
11.
IS.
13.
14.
15.
16.
17.
18.
19.
86.
at.
22.
83.
84.
as.
86.
87.
88.
89.
36.
31.
38.
33.
34.
35.
36.
37.
38.
39.
46.
41.
48.
43.
44.
45.
46.
47.
48.
49.
56.
St.
52.
S3.
                     HOHER CITY PLANT UNIT
                               8.4399996
                               8.5699997
                               8.7786665
                               a. 7260003
                               2.4eeeeeo
                               2.5500002
                               2.5100002
                               2.sa0666S
                               2.5966668
                               2.7966666
                               2.6666664
                               8.5799999
                               2.7799997
                               3.0406660
                               a
                               a
                                           3 CSA CEIt DATA
 .4799995
 .4766663
a.54eeeee
2.7166006
8.5966668
8.8868668
8.8899999
8.6999998
2.5699997
2.6999998
8.S966863
8. 6499996
8.4700003
                               a. 5600004
                               8. 5299997
                               a.46«aeee
                               8.4399996
                               8.3880»»1
                               8. 4381603
                               8.5966608
                               8. 5006666
                               a. 7866663
                               8.S666664
                               8. 6466663
                               2.8866661
                               a. 4899998
                               8.4699998
                               2.4899998
                               8.5866605
                               2.4399996
                               8.6666664
                               8.3966663
                               8.3566664
                               2.3866661
                               8.7399998
                               2.S00664*
                               8.6766601
                     54.
DAILV 2ND OUERLAPPING55.
                     56.
                     67.
                     58.
                     59.
                     66.
                     61.
                     62.
                     63.
                     64.
                     65.
                     66.
                     67.
                     68.
                     69.
                     76.
                     71.
                     72.
                     73.
                     74.
                     75.
                     76.
                     77.
                     78.
                     79.
                     86.
                     81.
                     82.
                     83.
                     84.
                     85.
                     86.
                     87.
                     88.
                     89.
                     96.
                     91.
                     92.
                     93.
                     94.
                     95.
                     96.
                     97.
                     98.
                     99.
                    166.
                    161.
                    162.
                    163.
                    164.
                    165.
                  ?
2.4799995
2.4499998
2.6099997
8.8000008
8.6400663
8.5299997
8.6199999
2.6499996
8.6366001
8.6766661
a.6000604
2.7560000
2.7100000
2.7899995
2.6306661
2.1666004
8.4499998
8.9300003
8.4206061
8.4706003
2.3900003
2.3299999
2.5299997
2.8100004
2.3299999
8.5000000
8.6300001
2.3999996
2.5799999
2.4799995
 .4499998
 .4200661
 .1066064
 .5166062
 .3199997
 .3999996
2.5560002
2.4499998
8.5799999
8.6060604
8.6499996
8.7399998
a.5400000
8.6599998
2.4399996
2.4200001
2.8906003
2.7666008
2.6899996
2.7266663
2.4799995
8.5166662

-------
  tee.               8.6599998
  1*7.               3.7399995
  tea.               8.5899997
  109.               2.6199999
  114.               2.4899998
  lit.               2.3999996
  I18.               2.4399996
  113.               2.6199999
  1H.               2.6499996
  US.               2.4306093
  116.               2.4399996
  117.               2.6400003
  118.               2.S699997
  119.               2.5408806
  180.               2.4499998
  181.               2.9099998
  182.               2.6099997
  183.               2.8599997
  184.               2.5606004
  185.               8.5299997
  126.               2.6099997
  127.               2.6000604
  128.               2.7760005
  129.               2.7100660
  136.               8.6499996
  131.               8.5900002
  138.               8.6766661
  133.               8.7799997
  134.               8.6466663
  135.               2.5460060
  136.               8.8699999
  137.               8.4766663
  138.               8.6466663
  139.               8.5566668
  146.               2.3466668
  141.               2.5699997
  142.               2.6666664
  143.               2.7299995
  144.               2.7266663
  14S.               2.7299995
  146.               2.6999998
  147.               8.6000004
  148.               2.7506666
  149.               2.4709063
  150.               2.5966662
  151.               2.6660004
  152.               2.4600000
  153.               2.3900003
  154.               2.4600000
  155.               2.4300003
  156.               8.6099997
  157.               2.3199997
T

-------
               t.     HOHER CITV PLANT UNIT 3 CSA DATA
               a.               8.3000002
               3.               2.2260003
               4.               8.3899999
               s.               2.0400000
               6.               1.8199997
               7.               2.1S99998
               8.               8.3299999
               9.               8.6099997
              It.               2.8299999
              ti.               i.7ioooao
              18.               1.9799995
              13.               8.5200005
              14.               3.3599997
              15.               8.3900003
              16.               1.9200001
              17.               8.1499996
              18.               8.2899995
              19.               8.8100000
              80.               8.1599998
              81.               8.8600008
              82.               8.4600000
              83.               8.5600004
              84.               8.7799997
,              85.               2.7900000
rv>            86.               8.7700005
N            87.               8.4399996
              88.               8.7100000
              89.               8.2100000
              30.               8.3100004
              31.               2.S699997
              38.               8.7900000
              33.               8.4099998
              34.               8.5600004
              35.               8.5500008
              36.               8.7500000
              37.               8.5900002
              38.               8.5699997
              39.               2.7799997
              40.               8.6300001
              41.               8.8299999
              48.               8.6800003
              43.               8.8199997
              44.               8.8199997
              45.               2.5799999
              46.               8.4799995
              47.               8.5500002
              48.               2.6700001
              49.               2.6999998
              50.               8.8899999
              51.               8.5200005
              58.               8.3899999
              S3.               8.7900000

-------