Fuel Processing Technology, 7 (1983) 11--22 11

Elsevier Scientific Publishing Company, Amsterdam -- Printed in The Netherlands

FORMULAS FOR CALCULATING THE CALORIFIC VALUE OF COAL

AND COAL CHARS: DEVELOPMENT, TESTS, AND USES

DAVID M. MASON and KIRAN N. GANDHI

Institute of Gas Technology, 3424 South State Street, Chicago, Illinois 60616 (U.S.A.)
(Received June 16th, 1980; accepted May 17th, 1982)

ABSTRACT

A new formula for calculating the calorific value of coal from its ultimate analysis
has been obtained by regression analysis of a data bank consisting of data on 775 samples
of U.S. coals of all ranks. It yielded zero average difference between observed and cal-
culated values and a standard deviation of 129 Btu/lb (300 kJ/kg). Neither average dif-
ference nor standard deviation varied much with rank of the coal. The Dulong, Mott--
Spooner, Boie, and Grummel--Davies formulae were also tested with the data but gave
substantially poorer results. The new formula has been substantiated with data from
other laboratories and with data on chars. The distribution of variance, attributable to
variability of mineral matter, variability of coal organic matter not related to rank, and
variance of laboratory determinations was estimated. Use of the formula to monitor
performance of a laboratory is illustrated, as is modification to obtain maximum preci-
sion in heat balance calculations on a coal conversion process.

INTRODUCTION

The gross calorific value, also called gross or upper heating value, of coal
is routinely determined by bomb calorimetry according to one of the two
ASTM Methods [1, 2]. (The misnomer "heat c o n t e n t " has recently appeared
in the literature as a s y n o n y m for calorific value or heating value. Instead,
"heat c o n t e n t " is a thermodynamics term equivalent to enthalpy [3] .) Be-
cause water produced from hydrogen in the coal by combustion is condensed,
the measurement yields the gross rather than the net calorific value. The
gross calorific value is of great importance for heat balance calculations in
the conversion of coal to other useful forms of fuel, as well as in its direct
use in combustion. The significance of the correlation of calorific value with
composition in ordinary fuel usage is shown by the development, as early as
1940, of some 9 different formulae for calculating calorific value from the
ultimate analysis, and of 11 formulae for calculating it from the proximate
analysis [4]. At least four additional ultimate analysis formulae have been
proposed in recent years [ 5--8]. The correlation is perhaps of even greater
importance for the rationalization and modeling of conversion processes now
being developed.
12

Our own work on this problem was carried out for a project on prepara-
tion of a "Coal Conversion Systems Technical Data Book", supported by
the U.S. Department of Energy and its predecessors.

TEST OF FORMULAS

A data bank (including experimental gross calorific values, ultimate anal-

yses, forms of sulfur, and rank) was established, from sample data published
by the Coal Research Section of Pennsylvania State University [9] and 681
samples analyzed by the Bureau of Mines [10--13]. The Penn State samples
were channel samples from whole seams [14]. The data bank covers all
ranks and a wide range of coal fields of the United States. The calorific
values ranged from 7900 Btu/lb (18,375 kJ/kg) for a lignite to 15,320 Btu/lb
(35,635 kJ/kg) for a medium volatile bituminous coal.
Four formulas were selected for testing. They are as follows:
Dulong [4]

Q = 145.44 (C) + 620.28 (H) + 40.5 (S) -- 77.54 (O) (1)

Boie [15]

Q = 151.2 (C) + 499.77 (H) + 45.0 (S) -- 47.7 (O) + 27.0 (N) (2)
Grummel--Davies [4, 16]

I654.3 (H) + 424.62 ] [(C)/3 + (H) -- (O)/8 + (S)/8] (3)

Q= (100 -- (A))
Mott--Spooner [4, 17]
Q = 144.54 (C) + 610.2 (H) + 40.5 (S) -- 62.46 (O) (o) < 15% (4a)
30.96 (O)
Q = 144.54 (C) + 610.2 (H) + 40.5 (S) - [ 65.88 1 (o)
(100 -- (A)) J
(O) > 15 (4b)
In the formulas above, Q is the calculated gross calorific value in Btu/lb
on the dry basis and (C), (H), (S), (O), (N), and (A) are the respective con-
tents of carbon, hydrogen, sulfur, oxygen, nitrogen, and ash in weight per-
cent, also on the dry basis.
For a fair test of the formulas on samples representing commercial coal
we set aside samples with more than 30% ash, leaving a total of 775 out of
the original 802 samples. Those remaining consisted of 40 anthracite, 406
bituminous, 180 sub-bituminous, and 149 lignite samples.
With a given formula, subtraction of the calculated calorific value from
the observed experimental value yields a residual for each sample. Statistics
of the residuals were calculated for each rank of coal and also with all ranks
pooled, as follows:
 13

Mean = (Qi observed -- Qi calculated) /n (5)

R o o t Mean Square (RM8) = (Qi observed -- Qi calculated) II

/n
(6)
Standard Deviation = I ~[RMS)2 -- (Mean)2]n/( n -- 1) I ~ J (7)

where n is the number of samples of a given rank or the total of all ranks, as
the case may be. The mean of the residuals is the bias of the calculated
values, and these two designations are used interchangeably in this paper.
Results from the four formulas are presented in Table 1; the means (column
3) and standard deviations (column 5) give the best description of the sets
of residuals; the latter also indicate the results that would be expected if the
formulas were modified by a correction term for each rank of coal. We also
calculated the standard deviations (column 6) t h a t are obtained if each
formula is modified by addition of a single correction term, namely the all-
ranks bias obtained from our data bank.
In addition to calculation with the formulas per se, we also calculated
calorific values by use of Given and Yarzab's modified Parr equation for
mineral matter content, and their corrections to obtain carbon, hydrogen,
sulfur, and oxygen on a mineral-matter-free basis [18]. This calculation re-
quires values for pyritic sulfur t h a t were n o t available for some of the sam-
ples. Results obtained with the modified Mott--Spooner equations on 646
samples on which pyritic sulfur contents were available are also shown in
Table 1; results from other formulas were also improved slightly, but the
Mott--Spooner equation gave the best results. Details of this calculation and
additional results are reported elsewhere [19].

DEVELOPMENT OF THE DATA BOOK FORMULA

The data bank was also subjected to a least squares regression analysis.
Carbon, hydrogen, sulfur, ash, and oxygen terms were significant; nitrogen
and cross and square terms were not. To reduce to a minimum the required
number of determinations, we adopted an oxygen-plus-nitrogen term in-
stead of an oxygen term. The resulting equation, which we refer to here as
the Data Book Equation, is as follows:

Q = 146.58 (C) + 568.78 (H) + 29.4 (S) -- 6.58 (A) -- 51.53 (O + N) (8a)

When 100 -- (C + H + S + A) is substituted for ( 0 + N), an equivalent form

is obtained:

Q = 198.11 (C) + 620.31 (H) + 80.93 (S) + 44.95 (A) -- 5153 (8b)
14

TABLE 1

Test o f f o r m u l a s for calculation o f calorific value

(i) (2) (3) (4) (5) (6)

Statistics of residuals (Btu/Ib)**
No. of Mean* Root Standard deviation
samples mean With rank Withall-ranks
square mean mean

Dulong
Anthracite 40 --123 155 97 144
Bituminous 406 --138 221 173 212
Sub-bituminous 180 128 212 170 222
Lignite 149 175 254 185 265
All R a n k s 775 - - 15 223 -- 222
Boie
Anthracite 40 --400 412 100 175
Bituminous 406 --248 279 129 129
Sub-bituminous 180 --207 249 138 146
Lignite 149 --298 329 138 144
All R a n k s 775 --256 291 -- 139
Grummel--Davies
Anthracite 40 80 132 106 163
Bituminous 406 --128 208 164 184
Sub-bituminous 180 47 167 161 185
Lignite 149 40 171 167 186
All R a n k s 775 - - 44 189 -- 183
Mot t--Spooner
Anthracite 40 - - 56 106 91 99
Bituminous 406 --134 197 144 149
Sub-bituminous 180 - - 31 151 149 162
Lignite 149 - - 85 169 147 147
All R a n k s 775 - - 96 178 -- 150
Mo t t--Spoo ner
modified Parr basis
All r a n k s 646 - - 42 138 -- 132
N ew formula
Anthracite 40 - - 14 93 92 --
Bituminous 406 10 124 123 --
Sub-bituminous 180 13 140 139 --
Lignite 149 - - 45 137 129 --
All ranks 775 0 -- -- 129

*Average o b s e r v e d value - - average calculated value = bias o f t h e calculated values.

* * 1 B t u / l b = 2.326 kJ/kg.

Results from this new formula a r e a l s o s h o w n i n T a b l e 1. T h e b i a s f o r d i f -

ferent ranks of coal ranges only from --45 Btu/lb (105 kJ/kg) on lignite to
13 Btu/lb (30 kJ/kg) on sub-bituminous coal and does not show a trend with
r a n k . T h e s t a n d a r d d e v i a t i o n ( o r s t a n d a r d e r r o r o f e s t i m a t e ) is s i g n i f i c a n t l y
 15

less than those of the other unmodified formulas, even after improving
these by a bias correction. (Note that eqn. (7) must be modified for use here
according to the appropriate loss of degress of freedom caused by evaluation
of the coefficients of eqn. (8a)). The new formula has a b o u t the same ac-
curacy as the Mott--Spooner with modified Parr corrections, b u t the latter
is more complicated and requires pyritic sulfur determination.
The effect of ash content on the accuracy and precision of the formula
was investigated, with the results shown in Table 2. For this test, the formula
was also applied to the 27 high-ash samples that had been set aside. It is
evident that the bias does not show a trend with ash content b u t that the
standard deviation does increase with ash content. The increase of the stan-
dard deviation with ash content can be attributed to the effects of differing
contributions of the mineral matter to determined carbon and hydrogen con-
tents, and differing heats of dehydration and reaction of the mineral matter.
A c o m p u t e r analysis of the data in which it was assumed that the variance is
linear with ash content indicated that the variance increases by (21.7 Btu/lb) 2
((50.5 kJ/kg) 2) per percent of ash, and that the standard deviation on ash-free
samples is 106 Btu/lb (247 kJ/kg).

TABLE 2

E f f e c t o f ash c o n t e n t o n t h e c a l c u l a t i o n o f calorific value with the n e w f o r m u l a

Ash c o n t e n t Number of samples Bias Standard deviation

wt % (Btu/lb)* (Btu/lb)

0--I0 394 6 113

10--20 320 --15 144
20--30 61 26 141
> 30 27 20 211

• 1 Btu/lb = 2.326 kJ/kg.

The contribution of pyritic sulfur to the variance was also investigated. A

regression analysis was performed in which a term for pyritic sulfur, added
to the existing formula, was evaluated. No decrease in standard deviation of
the 646 samples having pyritic sulfur values, or of those having 1% or more
of pyritic sulfur, was observed when the pyritic sulfur term was added.
It is of interest to determine h o w much of the remaining variance, (106
Btu/lb) 2 ((247 kJ/kg)2), can be attributed to the laboratory determinations.
The expected variance attributed to these can be estimated according to
Var (det) = Var Q + 198.112 Var (C) + 620.312 Var (H)

+ 80.932 Var S + 44.952 Var (A) (9)

where V a t Q, V a t (C), etc., are variances of the respective determinations.
16

Some precision data, reported in the Appendix, have become available from
the HYGAS ® program at IGT and can be used for the estimation. When ap-
plied in eqn. (9), the standard deviations of the repeatability determinations
(after bias corrections) from Table A-1 (in the Appendix) yield the value 64
Btu/lb (151 kJ/kg) as the expected standard deviation of the calculation
residuals. This represents 77% of the variance (73 Btu/lb) 2 ((170 kJ/kg)2),
found for a large set of HYGAS data discussed later in this paper. However,
it represents only 25% of the total variance, (129 Btu/lb) 2 ((300 kJ/kg)2), of
the original data set, or 36% of the variance estimated for mineral-matter-
free coal. As a source of the remaining variance of the latter, it is possible but
unlikely that the repeatability of the data bank determinations is significant-
ly worse than that of IGT determinations. The absence of a trend in bias with
rank of the coals (Table 1) indicates that the source of the unaccounted-for
variance is not related to rank. The most likely source is the effect of the
maceral composition of the coal.

TEST AND USES

Data for testing of the new formula were solicited from outside laborato-
ries. Results from three laboratories presented in Table 3 show good preci-
sion, but the large bias values suggest the presence of systematic error, or,
more likely, t h a t coals from a single mine or region exhibit a characteristic
bias.

TABLE 3

Tests of new calorific value formula at other laboratories

Laboratory Kind of coal No. of samples Bias Standard deviation

(Btu/lb)* (Btu/lb)

A Bituminous 42 --97 129

B Illinois Basin 78 68 113
C Sub-bituminous 40 --50 90

*1 Btu/lb = 2.326 kJ/kg.

Results from a fourth laboratory illustrate an important use of the formula.

The experimental data were obtained over a period of several years and were
furnished to us in the sequence in which they were obtained by the labora-
tory. We eliminated a few samples having more than 35% ash or less than 3%
oxygen, because the latter are likely to be chars. On the remaining data, the
root mean squares of the residuals obtained from consecutive sets of 50 sam-
ples are shown in Fig. 1. On the first 650 samples the mean residual was 32
Btu/lb (74 kJ/kg) and the standard deviation was 136 Btu/lb {316 kJ/kg), in
good agreement with our results on the original data bank. On subsequent
 17

5OO

z
400

~ 300
!iii
iiiii!i
i
$
w 200 ¸!
i~ ~i~iliii!
iiiiiiii!i!i
~i~i~i'i)
I!!I
2 0
.i i
0 ICO 200 300 400 500 600 70O 800 9OO I000

CONSECUTWE SAMPLE NUMBER

Fig. 1. Variability of the difference between observed and calculated calorific value in a
laboratory.

samples the results indicate a substantial deterioration in laboratory precision.

Thus, a control chart of this kind can serve as a monitor of laboratory per-
formance. Also, the difference between observed and calculated heating
values on an individual sample can be used by the laboratory supervisor as a
criterion of acceptability of the heating value and carbon--hydrogen deter-
minations. The difference is less sensitiveto errors of the sulfur and ash values.
Another important use of a heating value formula is in analysis and model-
ing of coal conversion processes. Data on heating value and composition of
samples of coal and char obtained at IGT under the H Y G A S pilot plant
program were analyzed for this purpose. In the H Y G A S process, non-agglom-
erating coals are dried but are not otherwise pretreated. Bituminous coals
are pretreated at temperatures of 750 ° to 800°F (400--425°C) to destroy
their agglomerating properties. The resulting product is referred to here as
"pre-treated coal" rather than "char". Samples referred to as "chars" are from
later intermediate gasificationstages or are spent (residue) char. Ash in the
spent char from the runs on bituminous coal averaged 36%, but ranged up
to about 85%. To augment the data from runs on sub-bituminous and lignite
coals, we have added some samples taken from streams that contain feed coal
in addition to char, such as the dust collected by a cyclone in the reactor
product gas stream. Results are presented in Table 4, together with those ob-
tained on the original data bank.
The most important criterion for use of a heating value formula in heat
balance calculations (for example, in a computer model) of a coal conversion
process is the bias or average difference between observed and calculated
values, because it shows h o w closely the formula represents the average
properties of the coal. O n the 294 samples of raw (untreated) bituminous
coals, the calculated values are, on the average, 18 Btu/Ib (42 kJ/kg) less
18

TABLE 4

Test on HYGAS routine samples of the new formula for calculation of calorific value

No. of Bias Standard deviation

samples (Btu/lb) a (Btu/lb) a

H Y G A S Data B a n k
B i t u m i n o u s coal b
Raw 294 18 71
Pre-treated coal 572 157 76
Chars
First stage hydrogasification 105 58 79
Second stage hydrogasification 106 2 106
Spent char 570 28 92
All char 781 28 94
S u b - b i t u m i n o u s coal c
Coal 49 --47 51
Chars 80 15 57
Mixtures of coal and char 66 15 53
Lignite d
Chars 80 34 77
Mixtures of coal and char 44 12 77
Original Data Bank
Anthacite 40 --14 93
Bituminous coal 406 10 124
Sub-bituminous coal 180 13 140
Lignite 149 --45 137
All samples 775 0 129

a l Btu/lb = 2.326 kJ/kg.

b F r o m Illinois No. 6 seam. About 5% of the raw coal samples and 17% of the others were
from runs on high-volatile B bituminous coal from Saline County, with the remainder
from runs on high-volatile C bituminous coal from Christian County.
CFrom the Rosebud Seam, Rosebud County, Montana.
d F r o m the Savage Mine, Richland County, Montana.

than the observed values. This differs by only 8 Btu/lb (18.6 kJ/kg) from the
value f o u n d on the 406 samples of bituminous coal in the original data bank.
The standard error shown for these coal samples is substantially less than
was f o u n d on the bituminous coals of the original data bank. The difference
may be a result of the limited range of source of the HYGAS samples: all
were from the Illinois No. 6 seam and 95% were from a single mine; the ash
c o n t e n t averaged 10.7%.
On pre-treated bituminous coal the calculated values are, on the average,
157 Btu/lb (365 kJ/kg} lower than the observed values. For better results,
a bias correction can be applied or a formula can be obtained by regression
analysis of the pre-treated coal data. The difference in bias between the
parent and pre-treated coals, 139 Btu/lb (323 kJ/kg), can be attributed to a
difference in structure (bonding); the formula has already taken into account
 19

differences in elemental composition. Such differences in structure include

effects of incorporation of oxygen in a different form from that ordinarily
present. In other processes the difference in bias may be greater or less,
depending on processing conditions such as temperature and presence or
absence of oxygen; in the HYGAS process the difference is reduced to about
40 Btu/lb (93 kJ/kg) at the stage where the temperature reaches a b o u t 1200°F
(650°C).
The calculated values on all chars from bituminous coals are, on the aver-
age, only 28 Btu/lb (65 kJ/kg) less than the observed values; the difference
is of a b o u t the same order for chars from sub-bituminous coal and lignite.
A more accurate formula could be obtained for the chars from the bituminous
coal, b u t the Data Book formula should be adequate for most practical pur-
poses; the accuracy should be judged according to unit weight of coal feed
rather than unit weight of char.

CONCLUSIONS

A new five-term formula for calculating the calorific value of coal from its
carbon, hydrogen, sulfur, and ash contents was obtained by regression anal-
ysis of data on 775 samples of U.S. coals of all ranks, most of which were
analyzed by the U.S. Bureau of Mines. The standard error of estimate of the
new formula was 129 Btu/lb (300 kJ/kg), compared with apparent standard
errors (root mean squares of differences) ranging from 178 to 229 Btu/lb
(414--533 kJ/kg) obtained on the same set of data from the Dulong, Boie,
Grummel--Davies, and Mott--Spooner formulas. The bias (average difference
of calculated from observed values) of the new formula varied with rank but
did not exhibit a trend with rank. The data indicate that there is little hope
of obtaining a more accurate formula based on ultimate analysis only.
The variance of the results with the new formula was analyzed and the
a m o u n t contributed by various sources was estimated. Approximately 32%
can be attributed to the effects of the variability of the mineral matter. The
variance contributed by variance of the experimental determinations, includ-
ing that of the calorific value, was estimated from repeatability data ob-
tained in IGT laboratories; if the repeatability of determinations in the
laboratories performing the analyses on the 775 samples of our data bank is
the same as that at IGT, then the variance from this source accounts for an
additional 26%. The remainder, 42%, is attributed to variability of the organic
matter unrelated to rank, such as maceral composition. Because the results
showed no trend in bias with rank of coal, little if any variance can be at-
tributed to rank-related properties of the coal.
Application of the formula to bituminous coal oxidatively pre-treated at
750 ° to 800°F (398°--427°C) to destroy agglomerating properties yields a
bias indicating that its heat of formation is higher than expected from ele-
mental and ash contents by a b o u t 140 Btu/lb (325 kJ/kg}; this is attributed
to difference in structure (bonding). The formula gives satisfactory results
20

on higher temperature HYGAS chars, and, with application of a bias correc-

tion, on pre-treated coal.
The formula is advantageous for use in heat balance calculations on conver-
sion processes and for monitoring test data on coal and char.

ACKNOWLEDGEMENT

The work reported here was conducted as part of a project, sponsored by

the U.S. Department of Energy, on preparation of a Coal Conversion Systems
Technical Data Book. Use of data obtained under the HYGAS program is
gratefully acknowledged.

REFERENCES

1 ASTM D 2015, 1981. Gross Calorific Value of Solid Fuel by the Adiabatic Bomb
Calorimeter. 1981 Annual Book of ASTM Standards, Part 26, American Society for
Testing and Materials, Philadelphia.
2 ASTM D 3286, 1981. Gross Calorific Value of Solid Fuel by the Isothermal-Jacket
Bomb Calorimeter. 1981 Annual Book of ASTM Standards, Part 26, American Society
for Testing and Materials, Philadelphia.
3 Rossini, F.D., 1950. Chemical Thermodynamics: 46. John Wiley, New York.
4 Selvig, W.A. and Gibson, F.H., 1945. Calorific Value of Coal. In: H.H. Lowry (Editor),
Chemistry of Coal Utilization, John Wiley, New York, Vol. 1, p. 139.
5 Subramanian, T.K., 1977. How to calculate Btu values of coal. Coal Age, 82: 153--
158.
6 Lloyd, W.G. and Francis, H.E., 1979. Personal communication.
7 Lloyd, W.G. and Davenport, D.A., 1980. Applying thermodynamics to fossil fuels.
Journal of Chemical Education, 57 : 56--60.
8 Neavel, R.C., Hippo, E.J., Smith, S.E. and Miller, R.N., 1980. Prepr. Pap. Am. Chem.
Soc., Div. Fuel Chem., 25(3): 246--257.
9 Spackman, W. et al., 1976. Evaluation and Development of Special Purposes Coals,
Final Report ERDA No. FE-0930-2, NTIS, Springfield, Virginia.
10 Swanson, V.E. et al., 1976. Collection, Chemical Analysis, and Evaluation of Coal
Samples in 1975. U.S. Department of the Interior, Geological Survey, Open File Report
76-468, Washington, D.C.
11 Gilmour, E.H. and Dahl, G.G., Jr., 1967. Montana Coal Analyses. Montana Bureau of
Mines and Geology, Special Publication 43, Butte, Montana.
12 Glass, G.B., 1975. Analyses and Measured Sections of 54 Wyoming Coal Samples
(collected in 1974). Geological Survey of Wyoming Report of Investigation No. 11,
Laramie, Wyoming.
13 Sondreal, E.A., Kube, W.R. and Elder, J.L., 1968. Analysis of the Northern Great
Plains Province lignites and their ash. U.S. Bureau of Mines Report of Investigation
No. 7158, Washington, D.C.
14 Institute of Gas Technology, 1978. Coal Conversion Systems Technical Data Book.
U.S. Department of Energy No. HCP/T2286-01, Superintendent of Documents,
Washington, D.C.
15 Boie, W., 1953. Fuel technology calculations. Energietechnik, 3: 309--316.
16 Grummel, E.S. and Davies, I.A., 1933. A new method of calculating the calorific value
of a fuel from its ultimate analysis. Fuel, 12." 199--203.
17 Mott, R.A. and Spooner, C.E., 1940. The calorific value of carbon in coal. Fuel, 19:
226--231,242--251.
 21

18 Given, P. and Yarzab, R.F., 1975. Problems and Solutions in the Use of Coal Analyses.
Technical Report No. 1, ERDA No. FE-0390-1, NTIS, Springfield, Virginia.
19 Institute of Gas Technology, 1978. Preparation of a Coal Conversion Systems Tech-
nical Data Book. ERDA No. FE-2286-32, NTIS, Springfield, Virginia.
20 ASTM D 2013, 1981. Preparing Coal Samples for Analysis. 1981 Annual Book of
ASTM Standards, Part 26, American Society for Testing and Materials, Philadelphia.

APPENDIX

Repeatability of Experimental Determinations at IGT

Sources o f variance t hat apply t o the problem were considered. The calorific
value and th e analytical determinations (carbon, hydrogen, sulfur, and ash)
are all run on an analysis sample of coal t hat has been ground finer than No.
60 sieve size [ 2 0 ] . Thus, the variance f r om sampling of the coarse sample
s u b mitted to the l a b o r a t o r y is n o t of concern. If the moisture c o n t e n t does
n o t change during all o f t he sample withdrawals for the various determina-
tions, no variance is c o n t r i b u t e d by the moisture determination; however,
if several days elapse between heating value and c a r b o n - - h y d r o g e n determina-
tions, a c o n t r i b u t i o n f r om this source is likely, either from a change in mois-
ture or f r o m the variance of its redetermination. Variance can also be con-
t r ib u ted by day-to-day Variations in e q u i p m e n t and operator; thus re-deter-
mination on the same day would n o t serve the purpose. Instead, our proce-
dure consisted of re-submitting f r om time to time a n u m b e r of analysis sam-
ples o f coal (in the same 4 oz (113 gm) bottles as originally sampled from)
for re-determination of moisture, heating value, carbon, hydrogen, sulfur,
and ash. Each r e p o r t e d value for carbon, hydrogen, ash, and calorific value is
the average o f t w o determinations run at the same time; for sulfur only one

TABLE A-1

Summary of IGT repeatability data

No. of Average Biasa Standard deviation (o)b

duplicates value
Before bias After bias
correction correction

Calorific Value 56 11,680 13 29 27

(Btu/Ib)c (wt %)
Carbon 41 64.07 0.037 0.25 0.24
Hydrogen 41 4.52 0.028 0.051 0.051
Sulfur 55 4.39 --0.011 0.084 0.086
Ash, SO3-free 40 16.77 0.037 0.14 0.14

aOriginal minus duplicate.

bof reported values, each the average of two determinations run at the same time, except
single determinations for sulfur.
Cl Btu = 2.326 kJ/kg.
22

determination is made. Completion of the duplicate analysis ranged from 9

to 46 days after completion of the original analysis. Slight average changes
in values from the original analysis to the duplicate, such as an average de-
crease in calorific value of 13 Btu/lb (30 kJ/kg), occurred; t h e s t a n d a r d
deviations were calculated both with and without correction for this effect.
The duplicate differences from this program were examined for outliers.
Three sulfur, one ash, and one calorific value, all with differences between
duplicates greater than 3.8 X 2 ~ o, were discarded. In addition, a calorific
value with a duplicate difference of 2.8 X 2 ~ a, and also having a difference
between observed and calculated values of 3.5 a, was discarded. The analysis
of the remaining data is presented in Table A-1.