CEMENT PROCESS ENGINEERING

VADE MECUM

2. STATISTICS

Rev. 2002

CEMENT PROCESS ENGINEERING

VADE-MECUM

SECTION 2 STATISTICS

Table of Contents
1. Descriptive Statistics ............................................................................... 2.1
1.1 Definitions .......................................................................................... 2.1
1.2 Basic .................................................................................................. 2.1
1.3 Normal Probability Distribution .......................................................... 2.1
1.4 Interval Estimation and Tables ............................................................ 2.2
2. Statistical Estimation Tests ..................................................................... 2.3
2.1 Generalities......................................................................................... 2.3
2.2 Test for the Equality of Two Variances ( 1 , 2 ) of two Normal
Population of Random Size, ( n1 , n 2 ) .................................................. 2.3
2.3 Fisher Distribution Table .................................................................... 2.4
3. Correlation Between Data Regression.................................................. 2.5
3.1 Generalities......................................................................................... 2.5
3.2 Least Squared Lines............................................................................ 2.5
4. Temporal/Regionalized Series (Variables) .............................................. 2.6
4.1 Stationnarity ....................................................................................... 2.6
4.2 Variogram .......................................................................................... 2.6
4.3 Raw Mix Control Tuning .................................................................... 2.9
5. Sampling ................................................................................................ 2.10
5.1 Golden Rules .................................................................................... 2.10
5.2 Fundamental Error (FE) .................................................................... 2.10
5.3 Minimum Representative Weight (MRW).......................................... 2.11
5.4 Estimation of the Maximum Particle Size .......................................... 2.11
5.5 Minimum Number of Observations.................................................... 2.12
5.6 Mechanical Sampling........................................................................ 2.12
5.7 Manual Sampling on Conveyor Belt .................................................. 2.13

Index - i
Rev. 2002

CEMENT PROCESS ENGINEERING

VADE-MECUM

SECTION 2 STATISTICS

1. Descriptive Statistics
1.1 Definitions

Statistics is the science of drawing conclusions about a population based on an analysis of sample data from
that population.
Population: values that can be taken by a variable.
Sample: drawing of n values of the variable taken from the population.
Random Variable = X = ( xi ) .
Probability Distribution = P ( xi ) . It describes the random variable probability of occurrence and is described
by its parameters. (Example: Normal distribution is described by and , see below).
Statistic = Any function of the sample data.
Estimator = An estimator of a parameter is a statistic, which corresponds to the parameter. For instance :
The sample mean ( x ) is the estimator of the actual population mean
The sample variance ( S 2 ) is the estimator of the actual population variance 2
Interval Estimation: An interval estimation of a parameter is the interval between 2 statistics that includes the
true value of the parameter with a given probability (1- ).

1.2 Basic

i =1

Arithmetical Mean = x =

2
Variance = S X
2
2
2
2
2 2
SX
+Y = S X + S Y and S aX = a S X

(x x)
n

Standard Deviation = S X =

a : Coefficient, X = ( xi ) , Y = ( yi ) : two series of independent values.

i =1

n 1

(x x )* ( y y)

Covariance = Average of the products of paired deviations: COV ( X ,Y ) = i =1

1.3 Normal Probability Distribution

The most often used probability distribution is the Normal probability distribution:
(

2
1 x x
)
2

dZ
1
=
e
dx 2
Central Limit Theorem

For a group of n independent sampling units drawn from a population of mean and variance 2 , the
sampling distribution of x =

x ,

1
n

i =1

x i is approximately Normal with mean and variance

2
. Said:
n

2.1
Rev. 2002

CEMENT PROCESS ENGINEERING

VADE-MECUM

SECTION 2 STATISTICS

1.4 Interval Estimation and Tables

a) If the Variance 2 is Known

The confidence interval, with a probability of (1- ), in which for any samples of the population with a given
unknown mean ( ) and known variance ( 2 ), the average x of the sample should range is given by:
x
2

x+

b) Normal Gauss Distribution Table

0.25
0.159
0.10
0.05
0.025

0.6745
1
1.28
1.64
1.96

0.0232
0.01
0.005
0.00135
0.001

2
2.32
2.57
3
3.09

Example:
Estimation of the true LHV mean ( ) of liquid waste fuel.
An n-size sample (n=100) of different waste fuel shipments gave a mean x = 5.5 MCal/kg. Standard deviation

of the waste fuel shipment population (considered infinite) is supposed to be 1Mcal/kg.

Then, according to the Central Limit Theorem, x follows a Normal distribution probability with a variance of

2
= 1 / 100 = 0.01 .
n
Thus, we are sure at 1 = 90% (then

[5.5 0.164 ,5.5 + 0.164] = [5.336 ,5.664] .

= 0.05 ) that the mean ( ) is between x 1.64 0.01 =

Remark:
If the population from which the sample is taken, is not infinite (lets say population size=800), then we have
to use a corrective factor of 1 n = 1 100
= 0.935 .
N
800
c) If the Variance 2 is Unknown and Sample Size n<30,

It has to be approximated by the variance S 2 of the sample. The Normal distribution is replaced by a t
distribution (Student distribution). The estimated interval, with a confidence of (1- ) and n-1 degree of
freedom, is given by:
S
S
x + t
x t
, n 1 n
, n 1 n
2

Example
With the data as above assuming S=1Mcal/kg, then with the same confidence (90%) and say 20 (21 samples)
degrees of freedom, ( ) is between: x 1.72 0.01 = [5.5 0.172 ,5.5 + 0.172] = [5.328 ,5.672] .

2.2
Rev. 2002

CEMENT PROCESS ENGINEERING

VADE-MECUM

SECTION 2 STATISTICS

d) Student Fisher Distribution Table

t 0.005 ,

t0.01,

t 0.025 ,

t 0.05 ,

t 0.10 ,

t 0.25 ,

t 0.45 ,

1
2
3
4
5
10
15
20
40
120

63.66
9.92
5.84
4.6
4.03
3.17
2.95
2.84
2.70
2.62

31.82
6.96
4.54
3.75
3.36
2.76
2.60
2.53
2.42
2.36

12.71
4.3
3.18
2.78
2.57
2.23
2.13
2.09
2.02
1.98

6.31
2.92
2.35
2.13
2.02
1.81
1.75
1.72
1.68
1.66

3.08
1.89
1.64
1.53
1.48
1.37
1.34
1.32
1.30
1.29

1
0.817
0.765
0.741
0.727
0.700
0.691
0.687
0.681
0.677

0.158
0.142
0.137
0.134
0.132
0.129
0.128
0.127
0.126
0.126

2. Statistical Estimation Tests

2.1 Generalities

A statistical hypothesis is a statement about the values of the parameters of a probability distribution.
Null hypothesis (H o ) : A = B , Alternative hypothesis (H 1 ) : A B .
To test a hypothesis, we take a random sample from the population under study, compute an appropriate test
statistic and then either reject or fail to reject ( H o ) with a risk of rejecting H o although H o is true.

2.2 Test for the Equality of Two Variances ( 1 , 2 ) of two Normal Population of Random
Size, ( n1 , n 2 )
(Excel Function FTEST)

Test Description

H o : 12 = 22 , H 1 : 12 22

We compute the statistic Fo =

We reject H o if Fo > F
2

Where

F
2

, n1 1, n2 1

S 12
S 22

, where F = Fisher Distribution

, n11, n2 1

or if Fo < F

and F

, n1 1, n2 1,
2

1 , n1 1, n2 1
2

denote the upper and lower

percentage points of the

F distribution with n1 1 and n2 1 degrees of freedom, respectively.

As the table for the F table gives only the upper tail points of the F, so to find F

, n1 1,n2 1
2

use: F

1 , n1 1, n2 1
2

1
F
2

we must

(be careful about n1 and n2 , which are inverted).

, n2 1, n1 1

2.3
Rev. 2002

CEMENT PROCESS ENGINEERING

VADE-MECUM

SECTION 2 STATISTICS

Example 1: Cement sampling:

We want to determine the best way of sampling cement. We can compare the variances of two sets of samples
collected at the mill discharge by two different ways. The H o hypothesis would be that there is no difference
between both ways of sampling (true variances equal). If the result says that according to the samples, there is
a difference, then the best sampling method could be the one with the lowest variance (one-sided alternative
hypothesis).
SSB measured:
#1 way: (4350, 4365, 4850, 4750, 4580, 4600, 4450, 4740),
#2 way: (4500, 4520, 4800, 4420, 4360, 4250, 4400, 4380)

H o : 12 = 22 , H 1 : 12 22 (two-sided alternative hypothesis)

(x 4586 )
8

After computing: x1 = 4586 , x 2 = 4454

S12 =

i =1

8 1

= 34796 , S 2 2 = 26598

34796
Fo =
= 1.308 < F0.05
= 4.99
26598
,8 1,8 1

and F.975 ,7 ,7 = ( F0.025,7 ,7 )

2
1

= ( 4.99 ) 1 = 0.20 <1.308

The test yields not to reject H o : the measurements dont allow us to conclude that #1 way of sampling is
significantly, with 5% confidence, different than #2 (even if S 1 > S 2 ). The excel function is FINV(0.025,7,7).

2.3 Fisher Distribution Table

(Calculated for the upper 2.5% of the F distribution (in our case, when = 5% )).
F(0.025, n1,n2)
1
1 647.79
2 799.48
3 864.15
4 899.60
5 921.83
6 937.11
7 948.20
8 956.64
9 963.28
10 968.63
15 984.87
20 993.08
25 998.09
30 1001.40
40 1005.60
50 1008.10
60 1009.79
70 1011.01
80 1011.91
90 1012.61
100 1013.16
200 1015.72

2
38.51
39.00
39.17
39.25
39.30
39.33
39.36
39.37
39.39
39.40
39.43
39.45
39.46
39.46
39.47
39.48
39.48
39.48
39.49
39.49
39.49
39.49

Ex: F(0.025,5,10)=4.24
3
4
5
6
17.44 12.22 10.01 8.81
16.04 10.65 8.43 7.26
15.44 9.98 7.76 6.60
15.10 9.60 7.39 6.23
14.88 9.36 7.15 5.99
14.73 9.20 6.98 5.82
14.62 9.07 6.85 5.70
14.54 8.98 6.76 5.60
14.47 8.90 6.68 5.52
14.42 8.84 6.62 5.46
14.25 8.66 6.43 5.27
14.17 8.56 6.33 5.17
14.12 8.50 6.27 5.11
14.08 8.46 6.23 5.07
14.04 8.41 6.18 5.01
14.01 8.38 6.14 4.98
13.99 8.36 6.12 4.96
13.98 8.35 6.11 4.94
13.97 8.33 6.10 4.93
13.96 8.33 6.09 4.92
13.96 8.32 6.08 4.92
13.93 8.29 6.05 4.88

7
8.07
6.54
5.89
5.52
5.29
5.12
4.99
4.90
4.82
4.76
4.57
4.47
4.40
4.36
4.31
4.28
4.25
4.24
4.23
4.22
4.21
4.18

8
7.57
6.06
5.42
5.05
4.82
4.65
4.53
4.43
4.36
4.30
4.10
4.00
3.94
3.89
3.84
3.81
3.78
3.77
3.76
3.75
3.74
3.70

9
7.21
5.71
5.08
4.72
4.48
4.32
4.20
4.10
4.03
3.96
3.77
3.67
3.60
3.56
3.51
3.47
3.45
3.43
3.42
3.41
3.40
3.37

10
6.94
5.46
4.83
4.47
4.24
4.07
3.95
3.85
3.78
3.72
3.52
3.42
3.35
3.31
3.26
3.22
3.20
3.18
3.17
3.16
3.15
3.12

15
6.20
4.77
4.15
3.80
3.58
3.41
3.29
3.20
3.12
3.06
2.86
2.76
2.69
2.64
2.59
2.55
2.52
2.51
2.49
2.48
2.47
2.44

20
5.87
4.46
3.86
3.51
3.29
3.13
3.01
2.91
2.84
2.77
2.57
2.46
2.40
2.35
2.29
2.25
2.22
2.20
2.19
2.18
2.17
2.13

25
5.69
4.29
3.69
3.35
3.13
2.97
2.85
2.75
2.68
2.61
2.41
2.30
2.23
2.18
2.12
2.08
2.05
2.03
2.02
2.01
2.00
1.95

30
5.57
4.18
3.59
3.25
3.03
2.87
2.75
2.65
2.57
2.51
2.31
2.20
2.12
2.07
2.01
1.97
1.94
1.92
1.90
1.89
1.88
1.84

40
5.42
4.05
3.46
3.13
2.90
2.74
2.62
2.53
2.45
2.39
2.18
2.07
1.99
1.94
1.88
1.83
1.80
1.78
1.76
1.75
1.74
1.69

50
5.34
3.97
3.39
3.05
2.83
2.67
2.55
2.46
2.38
2.32
2.11
1.99
1.92
1.87
1.80
1.75
1.72
1.70
1.68
1.67
1.66
1.60

60
5.29
3.93
3.34
3.01
2.79
2.63
2.51
2.41
2.33
2.27
2.06
1.94
1.87
1.82
1.74
1.70
1.67
1.64
1.63
1.61
1.60
1.54

70
5.25
3.89
3.31
2.97
2.75
2.59
2.47
2.38
2.30
2.24
2.03
1.91
1.83
1.78
1.71
1.66
1.63
1.60
1.59
1.57
1.56
1.50

80
5.22
3.86
3.28
2.95
2.73
2.57
2.45
2.35
2.28
2.21
2.00
1.88
1.81
1.75
1.68
1.63
1.60
1.57
1.55
1.54
1.53
1.47

90
5.20
3.84
3.26
2.93
2.71
2.55
2.43
2.34
2.26
2.19
1.98
1.86
1.79
1.73
1.66
1.61
1.58
1.55
1.53
1.52
1.50
1.44

100
5.18
3.83
3.25
2.92
2.70
2.54
2.42
2.32
2.24
2.18
1.97
1.85
1.77
1.71
1.64
1.59
1.56
1.53
1.51
1.50
1.48
1.42

200
5.10
3.76
3.18
2.85
2.63
2.47
2.35
2.26
2.18
2.11
1.90
1.78
1.70
1.64
1.56
1.51
1.47
1.45
1.42
1.41
1.39
1.32

2.4
Rev. 2002

CEMENT PROCESS ENGINEERING

VADE-MECUM

SECTION 2 STATISTICS

3. Correlation Between Data Regression

3.1 Generalities

Goal: express a dependant variable ( Y : ( y i )i =1ton ) as a function of one or a series of p independent variables
X j : ( X j = ( x j ,i ) j =1top ,i =1ton ) ).

Y E = b0 + b1 X 1 + .. + b p X p
Y E = estimated dependant variable, X j = independent variables.
We have n observation for each variable.

3.2 Least Squared Lines

The method minimizes the deviation E ( Ei ) between the points and the line.
x

(y y )
n

- SST = total sum of square of the variable of interest =

B1=y/x

(E
n

E=Y-YEst
YEst

- SSE = sum of square of errors =

Y
B0

i=1

i E

i =1

)2

- SSR = sum of square explained by the regression line: SST = SSR+SSE

X

We want to optimize SSR/SSE. Thus we test the hypothesis that the slope B1 equals 0:
H o : B1 = 0 , H 1 : B1 0 .
Under H o , the ratio (SSR/p)/(SSE/(n-p-1)) follows a Fisher distribution with p and n-p-1 degrees of freedom
(excel function FINV (, p, n-p-1)).
If F is high, then H o is rejected and with a certain significance , we assume the regression is significant.

Coefficient of Determination R2
The coefficient of determination R2=SSR/SST gives the proportion of variation in the dependent variable
( Y : ( y i )i =1 ton ) explained by the regression line.

The coefficient of correlation is defined by: r =sqrt (R2).

Example
H0: there is no correlation
n=5, p=1, SST=0.051+0.019, MSR=0.051/1=0.051, MSE=0.019/3=0.0063, F=0.051/0063=8.05,
.75
.7
.65

R2 = 0.051 / (0.051 + 0.019) = 0.73, r = 0.85

Critical F value ( = 0.025), F1,3,0.025 = 17.44 > 8.05
The ratio belongs to the F distribution
We cannot reject H0, the regression is not significant.

SO3

.6
.55
.5
.45
.4

Y = 2.077 - .032 * X; R^2 = .727

.35
42

43

44

45

46

47

48

49

50

51

52

CaO

2.5
Rev. 2002

CEMENT PROCESS ENGINEERING

VADE-MECUM

SECTION 2 STATISTICS

4. Temporal/Regionalized Series (Variables)

4.1 Stationnarity

The series X (t ) is stationary if its average X (t ) and its variance S 2 (t ) are constant (over time or over the
region of study) and if the covariance COV ( X (t ), X (t' )) does not depend on t and t'
difference (distance) t' t = t (= h ) .

but only on the

4.2 Variogram
a) Variogram Construction
A variogram is a plot of the average difference of a selected variable (C3S for example) between pairs of units
selected as a function of time, where the pairs are chosen in whole-number multiples (e.g. every minute, 2
minutes, 1 meter, 2 meters, ).

X (h ) =

N
x j x j+h

j =1

with :
- j : numbering of the samples value
- N: number of pairs of sample with a specific time or
spatial distance (=h) between values of a pair.

2 ( N 1)

Example:
The C3S values of kiln feed samples are:
Sample#
Time
C3S (%)

1
1:00
54.2

2
2:00
57.8

3
3:00
59.8

4
4:00
61.2

5
5:00
60.0

6
6:00
56.0

7
7:00
52.0

8
8:00
52.0

9
9:00
52.4

10
10:00
57.0

Then we can calculate the one-hour pair difference:

Pair#
Diff in pair
Square diff

1
3.6
12.96

2
2
4

3
1.4
1.96

4
-1.2
1.44

5
-4
16

6
-4
16

7
0
0

8
0.4
0.16

9
4.6
21.16

Sum
73.7

73.7
= 4.6
2 ( 9 1)
Two rules for variogram construction
Collect enough units (N) to get a statistical population (at least 30 samples for a short term experiment and 60
samples for a long term); the short term intends to define very precisely the random heterogeneity term (nugget
effect, refer below).
The number N should reach half the total amount of samples collected (N>n/2).

Then C 3 S ( 1 hour ) =

2.6
Rev. 2002

CEMENT PROCESS ENGINEERING

VADE-MECUM

SECTION 2 STATISTICS

b) Variogram Interpretation
Interpretation of the limit of variogram (h) when h increases
Whatever the variable is, beyond a certain value of h, the variable ceases to be
correlated with itself. It is because the phenomenon taking place has no longer
any memory of a past long gone (see case 2 and case 3 where the variable level
off at a sill generally equal to the variance of the variable).
This is true for all raw mix analyses, which are limited in terms of the values they
can take.
However, over a short period of time (a few hours), the signal may well drift.
(See graph below). In such a case, the variogram will tend to increase instead of

Signal is drifting

X(h)

stabilizing itself around x2 .

The "Nugget Effect"

Many variables, especially those obtained from data
measured with a dispersive method (analytical,
sampling errors, etc.), present a slight or marked
degree of strictly random variations from one value
to the next.
As a rule, a variable presenting a "smooth" graph (#
3) when plotted presents a low to non-existent
"nugget effect". (i.e. due to variability at a scale
smaller than the sampling distance).
A "noise" (# 1) presents all its variance as a "nugget
( x2 being

effect"
variance").

called

the

"nugget

2
2
x = xn

x (h)
#1

Nugget effect
t

h
x (h)

2
x

#2
Nugget effect
t

effect

h
x (h)

2
xn

2
x

#3
No nugget effect
t

Limitations in h value
If N values of X are available, shifts of more than N/2 should not be considered.
Regionalization and prediction
A very frequent pattern of variogram is
shown as below:
X (h )

2
x

2
xn
h
Area of ho
regionalization

The span of values of ho for which x (h) is below x2 is

called the "area of regionalization" or the range.
The value of the signal at time t + ho is in fact dependent
of all values taken by X between t and t + ho.
If all values xb x i +1 , xi + h +1 are known, then xi + h can
be predicted much better than by saying that it is
2
randomly distributed with a variance x .
In fact, the variance of the prediction, at its best, will be

close to

X
2

which is much smaller than x2 .

2.7
Rev. 2002

CEMENT PROCESS ENGINEERING

VADE-MECUM

Pseudo-periodicity
The periodic variations can be self-sustained (control
cycle, oscillator, etc.) or induced by a periodic
phenomenon (buckets of elevator are unevenly distributed,
correction interval of raw meal).
Even if the periodicity is blurred on the graph of the signal
by random noises or variations of the period, the
variogram will tend to underline.
The variogram will hit a maximum, above the total
variance x2 , for a shift h of exactly 1 period. Maximum
and minimum will repeat themselves and fade away as h
increases. The fading will be quick if the pseudo period
varies much but slow if the signal is truly periodic.

SECTION 2 STATISTICS

x (h)
2
2 x

Pseudo Periodic signal

t
1 Pseudo-Period
x (h)

Periodic signal

2.8
Rev. 2002

CEMENT PROCESS ENGINEERING

VADE-MECUM

SECTION 2 STATISTICS

4.3 Raw Mix Control Tuning

Correctogram is a simple statistics tool which can be used to determine whether over-control or under-control is
occuring in a control loop. For spot checking, a plot of the correctogram can be used.
Plot the cartesian coordinates (x, y) where:
x = values of control parameter set point, at time t
y = values of control parameter set point, at time t t

t is the sampling interval:

Example:
Time

C3S

SP C3S SP

2:00

64.1

60

4.1

4:00

58.5

60

-1.5

(4.1,-1.5)

6:00

58.9

60

-1.1

(-1.5,-1.1)

8:00

61.7

60

1.7

(-1.1,1.7)

10:00

56.7

58

-1.3

(1.7,-1.3)

(x , y)

4
3
2
1
0
-5

-4

-3

-2

-1

12:00

59.2

58

1.2

(-1.3,1.2)

-1

14:00

54.5

58

-3.5

(1.2,-3.5)

-2

16:00

60.8

58

2.8

(-3.5,2.8)

18:00

55.1

58

-2.9

(2.8,-2.9)

20:00

58.3

58

0.3

(-2.9,0.3)

22:00

59

58

1.0

(0.3,1.0)

-3
-4

SLOPE

-5

INTERPRETATION & CORRECTIVE ACTION

=0

Perfectly tuned control. All off-target values for the control parameter are due to random
variations (materials, feeder accuracy, etc.)
1 > slope > 0 Undercontrolling. Multiply gain by (1 + slope).
=1
No control taking place.
>1
Divergent control: gain value has wrong sign.
0 > slope > -1 Overcontrolling. Divide gain by (1 slope).
= -1
Overcontrolling is inducing a cycle with frequency = 2 x sampling interval. Divide gain by 2.
< -1
Divergent cycling due to severe overcontrolling. Divide gain by (1 slope).
The method is applicable to control response analysis in general.
It can be incorporated as an internal tuning device in a control algorithm.
Analyses of non linear control response can be performed by using polynomial fit rather than linear regression.

2.9
Rev. 2002

CEMENT PROCESS ENGINEERING

VADE-MECUM

SECTION 2 STATISTICS

5. Sampling
5.1 Golden Rules

The MRW.
The sampling method must allow every particle the same chance of being collected.

5.2 Fundamental Error (FE)

Calculation
This error can never be cancelled because it is intrinsic to the material. However, we want to collect the right
size (MRW) of the sample based on this Fundamental Error (P. Gys theory).
(1 )
2 (FE ) = C x d M 3 x
m
With:
d M : Top particle size (95% passing) in cm.
: sampling proportion (usually quite small, then 1- = 1)
m : sample weight in g.
C : Constant characterizing the material sampled, in g / cm 3

C = fcl g with
f = Particle shape factor. (= 0.5 usually, ranges between 0 and 1)
= 1 when cubic, = 0.2 when flat, = 0.5 when spheroidal
l = liberation factor [0 to 1]
= 0 if homogeneous, = 1 if particles completely distinct, = .001 for homogeneous raw mix, = .2
medium, = .3-8 heterogeneous
g = factor describing the particle size distribution

If we call size range the ratio d M / d m of the upper size limit d M : (about 5% oversize) to the lower size
limit d m : (about 5% undersize):
Large size range ( d M / d m > 4): g = 0.25, medium size range (4 to 2): g = 0.50, small size range (< 2): g
= 0.75, uniform size ( d M / d m = 1): g = 1.00

c = Mineralogical composition factor g / cm 3

c=

p
i

1 ai
. i
ai

ai + (1 ai ) ic

With:
- pi = proportion of material I in the mix (%)
- a i = concentration of the critical within the material I (%) in mass ( g of CaO / g of solid )
-

pi =

volumetric weight of the material i g / cm 3

- ic = volumetric weight of the initial in the material

Usually we take i = ic

2.10
Rev. 2002

CEMENT PROCESS ENGINEERING

VADE-MECUM

SECTION 2 STATISTICS

Example:
Mix is crushed at 12.5 mm of 75% lime and 25% clay, CaO is the critical
Sample weight = 50 kg.
f = 0.5
l = 0.3
lime
content = 52%, CaO clay content = 24%
CaO

CaO = 2.7 g / cm 3 , lime = 2.7 , clay = 2.7, g = 0.25

c = 0.75 x

1 0.52

x 2.7
0.52

1 0.24
x 2.7 =
0.24

+ 0.25 x

1.869 + 2.137 = 4.00 g / cm 3

Then: C = f l c g = 0.5 x 0.3 x 4.0 x 0.75 = 0.15 g / cm 3

(1.25 )3 x 0.15

= 2.4 .10 3 is the fundamental error standard deviation.

50 ,000
Then the 95% probability confidence interval 2 ( FE ) is 0.0048 and then CaO content confidence interval is:
052.( 1 2 ( FE )) = 0.52 0.048% CaO . (Considering that 1 1 )

And:

(FE ) =

5.3 Minimum Representative Weight (MRW)

a) Lafarge Corp Simplified Formula

MRW = 18. f . .

d3

( FE ) 2

In case of material encountered in cement plant, we usually have ( FE ) 2 <0.06.

Example 1 estimation of the MRW
For quarry crushed stone with:
f = 0.5 , = 2.6 g / cm 3 , d = 1.75cm , ( FE ) 2 = 0.01
MRW = 18 0.5 2.6 1.75 3 / 0.01 = 12.54 kg
b) Estimation of the MRW (P. Gys formula)

What is the MRW considering the previous lot C = 0.15 g / cm 3 ?

Passing 95% = d M = 4.0 cm with ( FE ) = 0.04 (be careful, it is a relative standard deviation), then,
MRW =

3
C .d M

( FE ) 2

0.15 x 4 3

(0.04 )2

= 6 kg

5.4 Estimation of the Maximum Particle Size

Assuming we want to sample a maximum of 5 kg sample with a tolerate standard deviation of = 0.04

Then:

3
dM

M 2
C

d M =3

5000 x 0.04 2
= 3.8 cm
0.15

a) Rule of Thumb:
Maximum Particle Size (mm)
Min sample Coal (ISO1988), kg
Min Sample Aggregate, ASTM D75, kg

10
0.6
10

20
0.8
25

30

40

60

80

50
3
100

60

75

90

120

150

175

ASTM for the aggregate industry is very safe.

2.11
Rev. 2002

CEMENT PROCESS ENGINEERING

VADE-MECUM

SECTION 2 STATISTICS

5.5 Minimum Number of Observations

Once the right size (MRW) of the sample is calculated, we want to determine how many samples (n) have to
be collected to have the acceptable knowledge (precision P) of the parameter ( X ) we are interested in, with an
afforded risk .
The larger the sample size, the closer we can expect the sample mean X to be to the population mean X .
Refer to the Central Limit Theorem above. The reliability of X as an estimate of X is measured by the
standard error of the mean which is simply the standard deviation of the sample mean.

Rule of thumb: n =

2X
2

where:
-

2X is the variance of the material stream and, 2 is the variance of the mean (the variability desired in
X

the result).

Remark
Each sample must have the MRW in order to have a right observation of the parameter that we want to have
estimated.
Example
The small-scale random heterogeneity of the raw mix, expressed in C3S variance, at mill outpout is 10, thus

2X =10.
We would like to decrease this random heterogeneity to 2, thus 2 =2,
X

Then to achieve this goal we have to sample 10/2=5 increments. Normally they have to be collected closely to
one another (e.g. 30 second interval).

5.6 Mechanical Sampling

(Sampling ratio 1/1000 to 1/10)
a) Cutter Width and Velocity Rules of Extraction Correctness

(for flow < 500 m 3 / h ).

d M = Maximum particle diameter
W = Actual Cutter opening
W0 = Minimum theoretical cutter width
b) First rule of Extraction Correctness
For d M > 3 mm : W Wo = 3 d M
For d M 3mm : W Wo = 10 mm
c) Second Rule of Extraction Correctness
Irrespective of d M , if the actual cutter width is W = n Wo (with n 1 ) then the cutter velocity V should
not exceed Von = (1 + n ) 0.3 m / s
Economical Optimum is : W = W0 and V = 0.6 m / s

2.12
Rev. 2002

CEMENT PROCESS ENGINEERING

VADE-MECUM

SECTION 2 STATISTICS

d) Interval of Time between Increment

No more than 5 minutes, usually every 30 seconds.
Make sure the number of increments making up the sample is in excess of 6 (a best is 30, ASTM 2234
(coal) recommends 15 increments for cleaned and 35 for uncleaned coal).

5.7 Manual Sampling on Conveyor Belt

a) When the Belt is Stopped
Sample enough material with regard to MRW.
Sample over all the width of the belt making sure to collect everything and perpendicular to the belt.
The length of sampling over the belt should be greater than the width of the belt.
Make-up the sample with several increments (more than 6 at least) to get the MRW.
b) When the Belt Keeps Running
Basic rule: extract a full cross-cut section of the flow stream, in several increment if necessary.
The manual sampling device width must be at least 2.5 times the bulk material top size.
Interval of time between increment.
- no more than 5 minutes, usually every 30 seconds.
- number of increments in excess of 6.

2.13
Rev. 2002