Support‎ > ‎Technical articles‎ > ‎

5 Ways to Improve Confidence in Geomet Samples

posted Jan 2, 2021, 2:36 PM by Adam Johnston   [ updated Jan 2, 2021, 2:39 PM ]
Mineral deposits are widely heterogeneous, and their mineralogical, physical and chemical characteristics vary throughout the deposit. Consequently, the metallurgical performance of the resource, such as throughput, recovery, or product quality, also varies. Geometallurgy looks to explain that performance variability using geological knowledge, reducing risk and identifying opportunities for improvement. 
Consequently, the samples used in metallurgical testwork need to address this variability. Use of samples with poor representativity in a geometallurgical program will lead to unrealistic predictions. Fatal flaws may be missed, or opportunities missed may result in excessive conservatism.

Deliberate human intent may also be at play, so a simple and transparent system for selecting and documenting samples is a valuable tool to demonstrate corporate compliance.

Describing sample representivity is a requirement to comply with standards, such as NI 43-101 technical reports, however many reports do not give even basic supporting information.
As a minimum, documentation should show what samples were used, where they are from, the selection criteria used, and how they compare to the deposit in question.

How does leadership know if the samples used to support predictions were representative or not?
This article gives a high level review of some basic principles that can support a systematic and transparent sample selection process. 

Note that this advice is worthless if not accompanied by expert knowledge in the geology and metallurgy of the resource.

1. Identify the Principal Features

First, the principal features to be considered should be identified jointly by the geologist and metallurgist. Features that often impact the metallurgical performance, include:
  • Geochemical analysis (valuable elements and contaminating elements),
  • Geological characteristics (mineralization, alterations, lithologies),
  • Geological domains, zones or specific locations.
  • Geotechnical characteristics such as rock quality, degradation, and rock structure

In an early stage project, it is important to keep an open mind about what may be important. It is common to base assumptions on the last project, but every deposit has its own secrets, and the mountain has no obligation to conform to your model.

A good way to identify possible principal features is to walk through the geological events that have occurred, including:
  • Pre-mineralization events,
  • Each mineralization event, and
  • Post-mineralization events
Geologists are usually most interested in the features that lead to new ore discoveries, and so it is important that the metallurgist assists in identifying features that may impact the metallurgical performance of the resource in question.

As process design testwork advances, the list of principal features should be reviewed and updated.

2. Select a Focus Case

Once the principal features are identified, the next step is to clarify what it is that the samples are to represent. 

For example, should the samples represent the entire deposit, the in-pit reserves, or only the oxide ore-type? Each of these is a different focus case.

The drilling and block model data sets usually include a lot of information outside the material to be sampled and tested. Representivity analysis needs to compare the samples selected to a subset focus case of the total information available. 
Data sets: Exploration > Deposit > Resource > Plant Feed > Ore-Type

The distribution of the features in the samples should resemble the distribution of the features in the focus case data set. 

An easy way to approach this is to work out what is plant feed, and what is waste. While plant feed usually cannot be called “ore” during the geometallurgical study, as it has not yet been demonstrated to be economically viable, Net Smelter Return (NSR) and Cut Off Grade (COG) estimates are good proxies for economic viability at this stage. Care should be taken to include dilution in the focus case, and not to filter the data at the COG. Mining COG is developed on a block by block basis, whereas metallurgical samples are taken from drillings that have much higher grade resolution. Taking samples only above the COG will mean that the characteristics of the mining dilution material will be missed.

3. Review the Samples

Get a collection of physical specimens or photos of each rock type feature for the deposit. Exploration geologists usually have a rock library on hand.

Visit the core shed if you can, but if that is not possible, ask to see the core photos. There is nothing that can replace the best practice of “see, touch, wet, hit, scratch the rock”. When you go to look at the samples, have all of the geological, geochemical, geotechnical, and mine plan information about each sample with you so that you can review those labels together with the physical sample.

Strip logs are a great way to organize your data for quick reference.

Mislabeled samples will cause problems when developing geometallurgical models, so have the features checked when samples are extracted and sent for testing. Confirming properties such as lithology or mineralogy by geochemistry is often helpful.

4. Quantify Representivity

Representivity can be defined as “the extent to which the sample displays the same characteristics as the resource it represents, in the test the sample is intended for.”

For example, comminution samples are representative if they have the same abrasion, competency and hardness as the rock they represent.

Metallurgical testing is done on a sparse number of samples (often less than 200 samples) compared to geochemical sampling (usually more than 50,000 samples). Therefore, Gy’s Theory of Sampling (TOS) approach cannot be applied, and typical geostatistical reporting methods are inadequate in explaining if a phase of metallurgical sampling was representative or not.
It is important that the samples selected cover the extremes of the principal features. Fatal flaws can often reside in the extremes.

Sample characteristics with respect to mine plan, min zones, rock types, structures, and minor element grades are just as important as checking the distribution of pay metal grades.
For numerical features a simple representation of the variability of elements in the deposit is by percentile curves, and the samples selected should be evenly distributed through the curve.
For categorical features, histograms that compare the focus case distribution to the sample population distribution are useful.

In TOS, representivity is assured during the sampling process.
In geometallurgy there is no way to take a representative sample using TOS. The term becomes subjective, and it is only known if the samples were representative once the resource is exploited, if at all. 
Representivity in geometallurgy is often accepted under the following conditions:
  • Principal feature distributions in sample and focus case populations are similar
  • Test results in each domain have continuous and normally distributed values
  • At least 30 samples in each significant domain
  • Spatial coverage throughout the deposit is complete (x, y, and z)
  • There is sufficient redundancy, no particular sample has high leverage over predictions
  • Mine blends have been tested for toxic or synergistic effects between ore types.
  • At least one sample per each 2 to 4 weeks of plant production over the life of mine

5. Use Cancha Geometallurgy Software

Cancha can be used to to structure, facilitate and automate points 1 to 4.
With the launch of Cancha 2.0, users now have access to the new geometallurgical sample selection, and representivity analysis functionality. This is a huge step forward for the industry, reducing sample selection time by over 95%. 

The following features in Cancha are used for sample selection:
  • Multiple focus case scenarios
  • Automated sample selection in seconds
  • Interactive review of samples between 3D, Logs, graphs and tables
  • Full documentation
  • Communication facilitated with 3D, strip logs, charts and data export
  • Distance to nearest sample calculated for each cell in blockmodel
  • Representivity reports on new, or existing sample sets
  • Proven on projects and operations with <10 to >1000 samples
No expertise in data science, programming, or advanced statistics required. Using a simple and intuitive interface, geologists, metallurgists, or other team members can select new samples and review sample representivity of existing samples in just a few clicks.

Contact us for more information about Cancha. 

Bonus 6. Name your samples systematically

“MET-01” is not acceptable! Cancha can help with naming samples too.