The business opportunities of automation technology are potentially limitless, but they will require radically new strategies and structures. Automation technologies will cause mining companies to rethink nearly everything they do, including actively nurturing a new range of skills in their future workforce. The field of machine learning will provide direction, strategy and tools.
Structures and inter-relationships between mining processes and mining data are being discovered through the use of new generation smart tools including augmented intelligence and machine learning.
Enterprise systems that incorporate machine learning approaches will drive the need for personnel who are skilled in more than collecting, analysing and reporting data. Interrogating data in the context of a mine’s operational model, and inferring connections between data and processes will be key to unlocking future productivity gains and establishing best practice.
Maptek’s new machine learning engine provides the ability to generate domain boundaries direct from drillhole sample data into a block model. This eliminates the need for time consuming manual digitisation and slow and cumbersome mathematical functions.
Embracing new technologies such as digital transformation and automation is a proven way to speed up the strategic planning cycle. Applied to resource modelling, it empowers miners who are looking to new strategies to exploit valuable resources, address issues such as stakeholder demands, talent shortages and risk management.
Machine learning fosters a more agile approach that delivers results with accuracy that compares favourably to traditional resource modelling methods. This allows resource geologists and analysts to review multiple projects in a much shorter time frame, such as to take advantage of commodity pricing, investment environment or takeover window. Spending months to generate models and test geological assumptions is guaranteed to result in lost acquisition/project opportunities.
The new Maptek approach of generating domain boundaries from drillhole sample data represents the best way to conduct geological modelling. A machine learning model that works on neural networks correlates geological database codes directly into a 3D block model, and then uses resulting domain codes to constrain grade estimation.
To date, computer-based modelling has provided advantages in speed and ease of use. However, it has traditionally replicated hand-drawn methods for building a model that portrays the best understanding of geological observations and facts. For example, sectional interpretations are joined into 3D solids, or surface contours are generated to build 2D boundary surfaces. Recently, advances have been made where simulation of categorical variables and implicit techniques have been used for semi-automatic generation of resource domains.
The deep learning domain determination is fast, taking a few minutes to assess and generate the results from thousands of drillholes. Sub-blocking within the resultant block model and measures of uncertainty are also incorporated into the deep learning modelling process, providing high levels of confidence in the final result.