Professional Documents
Culture Documents
Final Paper Sampling2018 Bandeira-Chieregati-Silva
Final Paper Sampling2018 Bandeira-Chieregati-Silva
bernardo.bandeira@mrn.com.br
2 Professor, Department of Mining and Petroleum Engineering, University of São Paulo, São Paulo,
Brazil, ana.chieregati@usp.br
3 Industrial Manager, Mineração Rio do Norte, Porto Trombetas, Brazil, jea.silva@mrn.com.br
ABSTRACT
The quality of sampling equipment can be verified by several statistical tests, whose theoretical
foundations and experimental procedures are described in international standards. These standards
detail procedures for checking the bias and the precision of sampling, which are mainly based on
classical statistics applied to random variables. However, when analysing variables such as ore
contents which are not strictly random but, on the contrary, are regionalized variables, the concepts
of Pierre Gy’s Theory of Sampling (TOS) are essential for a detailed understanding and the
minimisation of all sources of sampling errors. Describing the exhaustive work of optimising
sampling procedures, protocols, and equipment at Mineração Rio do Norte (MRN), a bauxite mine
in the North of Brazil, this paper presents the final certification of MRN’s port sampling station, using
the existing international standards and complementary work based on the TOS for sampling
aluminium ores.
1. INTRODUCTION
Mineração Rio do Norte (MRN) is located in the Porto Trombetas complex, in the municipality of
Oriximiná, west of Pará state, Brazil. MRN started its operations in 1974 as an association of national
and international companies with the exploring and commercializing bauxite, raw material for
aluminium. The operations consist of the ore extraction, processing, railroad transport, drying and
ship loading.
An extensive work of optimising sampling equipment and procedures at the port sampling station
(Figure 1) was performed in 2017 and 2018 for certification purposes. Following the principles of the
Theory of Sampling (Gy, 1992; Pitard, 1993) together with the methods described in international
standards for checking the bias of sampling, the samplers for particle size distribution, moisture and
chemical analysis were tested and validated. The following items describe the sampling methodology
and the results of the certification work in MRN's sampling station for both products supplied by the
company: the wet and the dry bauxite.
Figure 1: MRN’s port sampling station.
2. METHODOLOGY
According to the experimental methods of the Australian Standard AS 2806.6:2003, the results
obtained from the method to be checked ("Method B") are compared with the results of a reference
method ("Method A"), which is considered a method that produces practically non-biased results
from the technical and empirical viewpoints. In case there is no statistically significant difference
between the results obtained by Method B and those obtained by Method A, Method B can be
adopted as a routine method.
Based on the standards, the bias is estimated from the application of the unilateral t-test at a 5%
significance level, thus determining whether the difference between the results of the two methods
is due to random variations or whether the results are statistically different. The number of paired
data should not be less than 20 and it depends on the standard deviation of the differences based on
20 pairs of data and the value of the bias, , to be detected.
It is recommended that even after performing a series of experiments, the same experiments should
be repeated at regular intervals and whenever there is a change in ore quality. The experiment should
also be repeated when there is a change in sampling process or sampling equipment.
Method B is considered to be unbiased when |t0| < t. The value of t is related to the number of sample
pairs, k, and can be found in the standards, while t0 can be calculated according to the following
equation.
d
t0 = (1)
sd
k
Where d is the mean of the differences between sample pairs, sd is the standard deviation of the
differences and k is the number of sample pairs.
The number of sample pairs, k, is considered sufficient when the required number of pairs nr ≤ k. The
required number of pairs, nr, can be found in the standards, according to the normalized difference,
D:
D= (2)
sd
Where is the magnitude (in %) of the bias to be detected, and sd is the standard deviation of the
differences.
(3)
Where f is the form factor (0.5), is ore density (2.6), d95 is the diameter where 95% in mass of the
particles of the lot to be sampled pass (3/4” or 1.91 cm), sFSE is the maximum acceptable relative
standard deviation for the Fundamental Sampling Error (10% or 0.10 in the equation).
Following the standards for aluminium ore and respecting the minimum sample mass, 30 pairs of
samples were collected from the AM-4701 sampler and from the belt feeding the sampler. For the
moisture content bias test, the belt samples were taken as a reference for the calculations and the bias
to be detected varied from 0.4% to 1.5% (moisture content absolute value).
0,6799 d195
,865
Ms = = 909 g (wet-basis ore) (4)
s 2FSE
0,6347 d195
, 9164
Ms = = 877 g (dry-basis ore) (5)
s 2FSE
Where d95 is the diameter where 95% in mass of the particles of the lot to be sampled pass (3/4” or
1.91 cm), sFSE is the maximum acceptable relative standard deviation for the Fundamental Sampling
Error (5% or 0.05 in the equation).
The formulas presented are based on the calibration of the sampling constants (K and ) for the worst
case (SiO2), in order to guarantee the minimum sample mass for both the determination of available
alumina and of reactive silica. Figures 2 to 5 show the heterogeneity charts for wet-basis and dry-
basis ore, considering Al2O3 and SiO2.
Figure 2: Correlation between IHL and fragment size (Al2O3 - WET-BASIS ore).
Figure 3: Correlation between IHL and fragment size (Al2O3 - DRY-BASIS ore).
Figure 4: Correlation between IHL and fragment size (SiO2 - WET-BASIS ore).
Figure 5: Correlation between IHL and fragment size (SiO2 - DRY-BASIS ore).
The results show that, even with an average relative difference for moisture content of +6.14% in
favour of the official samples, the statistical test indicate that the AM-4701 sampler presents no
significant bias for determining moisture content, being therefore validated for moisture content
estimation. Test results also indicate that the number of sample pairs was sufficient to detect a bias
of 1.5% moisture.
3.2. Validation of the moisture analysis sampler: AM-4701 (dry-basis ore)
Table 4 shows the bias test results comparing the official samples with the belt samples for dry-basis
ore.
Table 4: Results of the bias test for AM-4701 (dry basis): official x belt (moisture).
Pair OFICIAL Absolute Relative
GRUBBS
index Sample ID OFICIAL CORREIA difference di di2 difference
B-A B-A Z OF Z COR
1 1 6.34 5.92 0.419 0.176 7.08% 0.693 0.062
2 2 6.37 6.24 0.132 0.017 2.11% 0.745 0.577
3 3 6.41 6.15 0.263 0.069 4.27% 0.815 0.394
4 4 6.51 7.18 -0.673 0.453 -9.37% 0.983 2.474
5 5 5.55 5.55 0.009 0.000 0.17% 0.680 0.825
6 6 6.74 5.81 0.929 0.863 16.00% 1.381 0.293
7 7 5.92 5.79 0.130 0.017 2.25% 0.042 0.331
8 8 5.99 5.22 0.764 0.584 14.63% 0.070 1.477
9 9 6.01 5.15 0.859 0.738 16.69% 0.107 1.625
10 10 5.34 5.80 -0.462 0.213 -7.96% 1.055 0.310
11 11 5.39 6.45 -1.054 1.110 -16.35% 0.963 0.989
12 12 6.31 6.65 -0.337 0.113 -5.06% 0.635 1.394
13 13 5.31 5.59 -0.287 0.083 -5.14% 1.113 0.728
14 14 5.36 5.57 -0.214 0.046 -3.85% 1.020 0.768
15 15 6.58 6.07 0.508 0.258 8.37% 1.102 0.233
16 16 6.05 5.82 0.229 0.053 3.93% 0.188 0.263
17 17 5.79 5.96 -0.170 0.029 -2.84% 0.265 0.015
18 18 5.81 5.17 0.638 0.407 12.34% 0.237 1.579
19 19 6.36 6.58 -0.216 0.046 -3.28% 0.730 1.259
20 20 6.35 7.01 -0.660 0.435 -9.42% 0.701 2.121
21 21 6.30 6.11 0.190 0.036 3.12% 0.624 0.319
22 22 6.12 6.30 -0.182 0.033 -2.90% 0.303 0.699
23 23 5.05 5.27 -0.228 0.052 -4.32% 1.566 1.372
24 24 5.97 6.00 -0.025 0.001 -0.42% 0.050 0.089
25 25 4.87 5.87 -0.997 0.994 -16.99% 1.872 0.177
26 26 4.36 5.52 -1.160 1.346 -21.03% 2.763 0.880
27 27 6.69 5.67 1.020 1.040 18.00% 1.291 0.580
28 28 6.53 5.97 0.555 0.308 9.30% 1.009 0.030
29 29 6.09 6.12 -0.037 0.001 -0.60% 0.243 0.336
30 30 5.89 6.13 -0.234 0.055 -3.82% 0.095 0.343
Mean dbar 5.95 5.95 -0.010 0.16%
Variance 0.33 0.25 0.330 0.98%
k 30
Sum -0.289 9.578
SSd 9.575
sd 0.575
Std Deviation 0.574 0.496 0.575 9.9%
t0 -0.092
t 30 pairs 1.699
UNBIASED? YES |t 0 |<t
(%) bias to be detected 0.40
D 0.70
nr # of pairs required 24
SUFFICIENT? YES nr<k
The statistical test indicate that the AM-4701 sampler presents no significant bias for determining
moisture content considering the official kiln-dried samples (average relative bias of +0.16%), being
therefore validated for moisture content estimation. The test results also indicate that the number of
sample pairs was sufficient to detect a bias of 0.4% moisture.
Tables 5 and 6 show, for available alumina and reactive silica respectively, a summary of the bias test
results comparing samples from the AM-4702 sampler with those from the CT-4707 belt.
Table 5: Results of the bias test for available alumina: AM-4702 x belt (wet basis).
The results of the statistical test indicate that the AM-4702 sampler presents no significant bias for the
determination of the alumina content, considering the comparison with the samples from the CT-
4707 belt. Test results also indicate that the number of sample pairs was sufficient to detect a bias of
2.5% available alumina.
Table 6: Results of bias test for reactive silica: AM-4702 x belt (wet basis).
The results of the statistical tests indicate that the AM-4702 sampler presents no significant bias for
the determination of reactive silica content when considering the comparison with the samples from
the CT-4707 belt. Test results also indicate that the number of sample pairs was sufficient to detect a
bias of 1.0% reactive silica.
Table 7: Results of bias test for available alumina: AM-4702 x belt (dry basis).
Table 8: Results of the bias test for reactive silica: AM-4702 x Belt (dry basis).
The results of statistical tests indicate that the AM-4702 sampler presents no significant bias for the
determination of reactive silica content when considering the comparison both with CT-4707 belt
samples, and is, therefore, validated with respect to the collection of representative samples for
reactive silica content analysis. Test results also indicate that the number of sample pairs was
sufficient to detect a bias of 0.8% reactive silica.
The equivalent particle size distributions and the very low relative errors for grade estimation
indicate absence of significant bias of the samples collected by the sampler COR-4702, considered
validated concerning the collection of samples for particle size distribution analysis for both wet-
basis ore and dry-basis ore.
Finally, it is worth noting that the relative errors of the COR-4702 in estimating available alumina
and reactive silica contents, considering the weighted average calculated based on the mass retained
in each fraction, is very low, as per Figure 9, where the first 2 columns represent the wet ore and the
last 2 columns represent the dry ore.
Table 9: Content relative errors for COR-4702 considering Al2O3 and SiO2 weighted averages.
4. CONCLUSIONS
The results of the sampling campaign for content determination indicated that the AM-4702 sampler
is validated, presenting no significant bias for the determination of available alumina and reactive
silica contents for both wet-basis ore and dry-basis ore.
The results of the sampling campaign for moisture determination indicated that the AM-4701
sampler is validated, presenting no significant bias for moisture content determination for both wet-
and dry-basis ore. It should be noted that the results are only valid for the total sample, that is, the
sample collected by the sampler without cone and quartering.
In relation to the sampler for size distribution analysis, COR-4702, the results showed that it tends to
lose some of the +3" particles. It was found that the likely reason for this loss is the breaking of the
fragments collected by the sampler when falling in the box located in a lower level. Except for the +3"
fraction, which represents less than 1% of the total sample mass, the particle size distributions of the
belt and COR-4702 samples presented no significant differences.
Based on the results presented in this paper and on the international standards AS 2806.6-2003 (2003)
and ISO 10226:1991 (1991), the port sampling tower of Mineração Rio do Norte is certified and its
sampling equipment are validated for both wet- and dry-basis ores.
Finally, in order to meet TOS principles, the authors recommend that both samplers (AM-4701 and
AM-4702) should be adjusted periodically, so that the bucket door is fully opened after cutting,
preventing the material to get stuck inside the bucket.
REFERENCES
AS 2806.6-2003 (2003). Part 6: Methods for checking the bias of sampling. In: Aluminium ores –
Sampling. Originated as AS 2806.6—1994. Second edition 2003. Standards Australia International,
Sydney, 14 p.
GY, P.M. 1992. Sampling of heterogeneous and dynamic material systems: theories of heterogeneity,
sampling and homogenizing. Elsevier, Amsterdam, 653 p.
ISO 10226:1991 (1991). Aluminium ores – Experimental methods for checking the bias of sampling.
International Organization for Standardization, Genebra, 24 p.
PITARD, F.F. 1993. Pierre Gy’s sampling theory and sampling practice: heterogeneity, sampling
correctness, and statistical process control. 2 nd ed., CRC Press, Boca Raton, Florida, 488 p.