Services

Terra Vision offers ore sorting evaluation services for all the phases of your project.  Please contact us directly or request information or a quotation for Sorting Amenabilty Studies through SGS Minerals Services in Lakefield, Ontario.

Supervised Sample Characterization

A supervised sample characterization is a  initial investigation to determine if there are color, texture, magnetic susceptibility, conductivity or x-ray transmission characteristics that can be used by an ore sorter to separate ore into distinct classes.

An expert (client or mineralogical characerization service) separates the sample into the desired classes. For example high-grade, grade1, grade2, low-grade, contaminants, waste, etc.  Images of the ore are acquired in a controlled lighting environment. Conductivity and magnetic susceptibility data can also be acquired.  The images are analyzed and descriptors such as color and texture are quantified. Finally the descriptors are analyzed to determine if there are patterns that can be used to differentiate the client-defined classes.

The results of characterization can be used to configure a full scale ore sorter for bulk sample tests.

Unsupervised Sample Classification

Unsupervised sample classification is a powerful tool to classify rocks into groups (or classes) with similar characteristics.  The advantage of this method is that classes of rocks can be created based on patterns that are not easily recognized by a manual inspection of the data.

Terra Vision uses unsupervised classification methods to generate virtual sorts based on the x-ray transmission, conductivity or magnetic susceptibility data as well as the image descriptors, such as texture, color, brightness, size and shape .  A virtual sort is a set of classes and the membership of each sample to each class. In other words, it describes how rock samples are related to one another, and how they can be classified together into groups of similar visual, conductivity and magnetic characteristics. The classes are generated through a combination of statistical and neural net data analysis methods.  Several virtual sorts can be generated from the same data set.

If assays are available for all of the individual rocks then this information can be combined with the virtual sort results to determine which sorting algorithm is optimal.  Alternatively, an expert can determine which virtual sort  results in the best classification and composite assays can be done for each class of the virtual sort. The results of this test can be used to configure a full scale ore sorter for bulk sample tests.

Bulk sample testing

Full scale sorting equipment is available for testing the throughput and recoveries of the ore.  Ideally the the sorting systems can be configured based on the results of previous supervised and unsupervised sample characterization tests.

Bulk samples varying from 250 kg to 10 tonnes can be accommodated.

Example Results

An example of the results of a sorting  study are shown in Table 1:Classes, Recovery, Mass, Grades. Figure 1:Sorter Heads and Sorter Tails vs % Mass Sorted represents this same information in a different format.

The results presented in Table 1 and Figure 1 are from actual run of mill , after the primary crusher, in an operating gold mine in South Africa. Although these results were generated from an unsupervised classification the information from the other types of tests is presented in a similar fashion.

The interpretation of the results presented in these two charts is fairly straightforward. For example, the line highlighted in yellow (class 16) shows that a sorter configured to reject all classes of ore below class 16 would only accept 53.27% of the total mass of ore.  The sorter heads would be 6.65 g/tonne and the sorter tails would have an average grade of 0.7 g/tonne. This case demonstrates that if the cost of processing gold ore at 0.7 g/tonne is greater than the final value of gold, then the mill could effectively be 42.63% smaller.