6
TIME AND COST SAVINGS
The methodology presented enables significant time savings
which translates to cost savings. It takes time to write code
in Python initially to standardize the procedure, but once
it is completed the process is efficient and cuts down time
from an average of 1200 seconds per drillhole for data clas-
sification and 12 hours for handpicking samples to around
60 seconds per drillhole for higher-quality data classifica-
tion and 8 hours for handpicking samples from the clas-
sified dataset. Initial data processing might differ for each
dataset, as multiple files may have to be combined and null
rows may have to be dealt with, but after that a standard
procedure is followed in which the cleaned database is fed
into Python functions to obtain the clustered database from
which samples are selected. The time savings exponentially
increase as the number of drill holes increases. This allows
for multiple iterations of sample selection quickly if needs
change during the study.
CONCLUSIONS
Selecting samples which are representative of the ore body
for metallurgical test work is an important part of a geo-
metallurgical study. A streamlined and standardized sample
selection methodology utilizing modern-day tools such as
Python is presented in this paper. The methodology was
used in a pre-feasibility study of a copper mine to select 40
samples with a minimum mass of 20 kg for flotation test
work. It resulted in grouping sections of drillhole lengths
with similar characteristics and elegantly selecting samples
with sufficient mass from those groups. Automating the
process resulted in significant time savings from an aver-
age of 1200 seconds per drillhole to around 60 seconds per
drillhole for data classification and from 12 hours to 8 hours
for handpicking samples from the classified dataset. There
is potential to use supervised machine learning techniques
in the clustering exercise of the methodology to capture
outliers more accurately, but cost and skill limitations may
have to be overcome as training the models requires time
and expertise. As with any sample selection program, even-
tually, the test work results dictate if the samples selected
resulted in the geometallurgy program’s desired outcome,
but this methodology allows multiple iterations quickly.
REFERENCES
[1] Potakey, N. E., &Ortiz, J. M. Defining geologi-
cal units using geochemical data and unsupervised
machine learning. Predictive Geometallurgy and
Geostatistics Lab, 47.
[2] Michaux, S., &O’Connor, L. (2020). How to set up
and develop a geometallurgical program. Geological
Survey of Finland.
[3] Dominy, S. C., O’Connor, L., Glass, H. J., Purevgerel,
S., &Xie, Y. (2018). Towards representative metal-
lurgical sampling and gold recovery testwork pro-
grammes. Minerals, 8(5), 193.
[4] Jansson, N. F., Allen, R. L., Skogsmo, G., &Tavakoli,
S. (2022). Principal component analysis and
K-means clustering as tools during exploration for Zn
skarn deposits and industrial carbonates, Sala area,
Sweden. Journal of Geochemical Exploration, 233,
106909.
[5] Lishchuk, V., Koch, P. H., Ghorbani, Y., &Butcher,
A. R. (2020). Towards integrated geometallurgical
approach: Critical review of current practices and
future trends. Minerals Engineering, 145, 106072.
Figure 9 .Comparing total time spent for data classification
using the old methodology and the new methodology
TIME AND COST SAVINGS
The methodology presented enables significant time savings
which translates to cost savings. It takes time to write code
in Python initially to standardize the procedure, but once
it is completed the process is efficient and cuts down time
from an average of 1200 seconds per drillhole for data clas-
sification and 12 hours for handpicking samples to around
60 seconds per drillhole for higher-quality data classifica-
tion and 8 hours for handpicking samples from the clas-
sified dataset. Initial data processing might differ for each
dataset, as multiple files may have to be combined and null
rows may have to be dealt with, but after that a standard
procedure is followed in which the cleaned database is fed
into Python functions to obtain the clustered database from
which samples are selected. The time savings exponentially
increase as the number of drill holes increases. This allows
for multiple iterations of sample selection quickly if needs
change during the study.
CONCLUSIONS
Selecting samples which are representative of the ore body
for metallurgical test work is an important part of a geo-
metallurgical study. A streamlined and standardized sample
selection methodology utilizing modern-day tools such as
Python is presented in this paper. The methodology was
used in a pre-feasibility study of a copper mine to select 40
samples with a minimum mass of 20 kg for flotation test
work. It resulted in grouping sections of drillhole lengths
with similar characteristics and elegantly selecting samples
with sufficient mass from those groups. Automating the
process resulted in significant time savings from an aver-
age of 1200 seconds per drillhole to around 60 seconds per
drillhole for data classification and from 12 hours to 8 hours
for handpicking samples from the classified dataset. There
is potential to use supervised machine learning techniques
in the clustering exercise of the methodology to capture
outliers more accurately, but cost and skill limitations may
have to be overcome as training the models requires time
and expertise. As with any sample selection program, even-
tually, the test work results dictate if the samples selected
resulted in the geometallurgy program’s desired outcome,
but this methodology allows multiple iterations quickly.
REFERENCES
[1] Potakey, N. E., &Ortiz, J. M. Defining geologi-
cal units using geochemical data and unsupervised
machine learning. Predictive Geometallurgy and
Geostatistics Lab, 47.
[2] Michaux, S., &O’Connor, L. (2020). How to set up
and develop a geometallurgical program. Geological
Survey of Finland.
[3] Dominy, S. C., O’Connor, L., Glass, H. J., Purevgerel,
S., &Xie, Y. (2018). Towards representative metal-
lurgical sampling and gold recovery testwork pro-
grammes. Minerals, 8(5), 193.
[4] Jansson, N. F., Allen, R. L., Skogsmo, G., &Tavakoli,
S. (2022). Principal component analysis and
K-means clustering as tools during exploration for Zn
skarn deposits and industrial carbonates, Sala area,
Sweden. Journal of Geochemical Exploration, 233,
106909.
[5] Lishchuk, V., Koch, P. H., Ghorbani, Y., &Butcher,
A. R. (2020). Towards integrated geometallurgical
approach: Critical review of current practices and
future trends. Minerals Engineering, 145, 106072.
Figure 9 .Comparing total time spent for data classification
using the old methodology and the new methodology