Knowledge systems and the mine of the future

Can the mining industry achieve its data-driven goals?

By Rudy Moctezuma

Superior productivity, efficiency, safety and ESG excellence: these are what today’s mines strive for. To realize these goals, they recognize their data plays a crucial role and that their improvements are limited by the extent of their capability to extract knowledge from their data.

To tackle this difficult task, they employ a wide range of commercial and in-house databases and platforms, designed to centralize, store, handle, process and analyze data.

But are they truly effective? And are they enough for the mine of the future?

Current landscape of data in mining
Today’s mining solutions generally use many or all the following locations, formats and attribution methodologies:

Location of data storage is a mix of local directories (usually multiple copies on the same machine in different states), local database instances, shared server drives and directories, shared database instances, SharePoint, ERPs and more.

Attribution is accomplished with filenames, file paths, internally to the file format or a set of database tables. There may be some basic tagging system in place, but it is usually up to the user to maintain.

Licensing restrictions range from text files and databases (which are generally unlicensed) to proprietary binary files accessible through an API or ‘data dump’ executable (generally licensed).

Data definition is interesting in that even if you have a text file with the data you want, it does not mean you have a description of what the data means. A good example is the use of an integer value to correspond to a specific characteristic (e.g., the number 1 = horizontally rotated model, you can read the 1, but do you know what that 1 means?).

Access methods vary widely. Most solutions touting their data as ‘open’ still require a specific tool for access. Data text files are open by their nature (i.e., I can open a text file up in Notepad) but proprietary binary files often need an API, dumping utility (which results in a text version) or you need to open the tool (which can be licensed), find the data you want, then export it into the desired format.

Formats are a mix of proprietary binary files, a variety of text (txt, ini, csv, xml, json, etc.), databases, Excel spreadsheets and others. Varying formats make it difficult to integrate data from multiple sources.

Current landscape of data technologies
In today’s digital landscape, organizations face a wide array of data management challenges due to the increasing volume, variety, and complexity of data – and all the various apps and users who need to access that data. Top among those challenges is the need to select the right data architecture and supporting technologies to meet evolving business needs and data requirements while ensuring data quality, security, and more.

Some of the most commonly known and used data technologies are data warehouses and data lakes, which are mostly repository systems, the former for structured data and the latter for raw data. While data warehouses support complex queries and business intelligence applications, they are costly and very rigid. Data lakes provide more flexibility and volume-scalability, but face challenges in data quality, governance and performance.

Next, we have Database Management Systems (DBMS), which are software systems that provide a standard method to store and organize data. While they can provide robust and reliable ways to manage structured data, they come with limitations, such as requiring substantial hardware resources, difficulties in scaling to very large volumes of data, and potential performance issues if running complex queries with large datasets.

Data mesh and data fabric are two more modern approaches to managing data and making it accessible. Data mesh focuses on decentralization and treating data as a product, while data fabric emphasizes a unified, intelligent layer for seamless data integration and management. Some organizations combine the two approaches. But they’re still left with the challenge of turning the huge volume of data into usable insights.

It can be safely said that choosing and implementing the right technologies can be a challenging task. Key considerations include flexibility, scalability, future-proofing, governance and seamless integration capabilities, as well as alignment with performance demands, resource allocation and technical expertise. These challenges are particularly significant as most organizations utilize a combination of various data technologies.

Ontologies and knowledge graphs
Some technology forward organizations, such as Google, Microsoft, Walmart and IBM Watson Health, use Ontology and Knowledge Graphs to organize their information in a way that they can understand the relationships between different concepts and entities.

Ontologies have proven to be powerful solutions to represent domain knowledge and integrate data from different sources. They enable a knowledge graph representation of data assets for different business domains.

In layperson’s terms, it helps organize data assets in a meaningful way by adding relationships and context. This way, when new data comes in, its effect on all related assets is felt, and users can easily see new insights.

As an example, consider a shift supervisor in load and haul operations. They are responsible for meeting the shift target: tons of material moved. KPIs are tied directly to that target, and the focus will be on ensuring the equipment is operating and assigned as efficiently as possible to move as much material as possible.

However, the focus is not on the material itself, the quality of that material, or its contents in ore and waste. Therefore, all decisions are based on moving material without considering material grade or quality. This causes multiple issues in mine operations, with instances in which waste material is mistakenly moved into ore piles and vice versa, causing a severe impact on mineral recovery and production.

A solution to this could include modified targets and KPIs to include a component that tracks material quality/composition. However, the supervisor can only consider this new component in their decision-making process if they have information on material composition and, therefore, understand how their decisions can affect the ore quality feeding into the plant. Using the knowledge graph, access to this needed knowledge becomes feasible.

Knowledge systems are the answer
Knowledge systems, such as SourceOne Enterprise Knowledge Performance System (EKPS) from Eclipse Mining Technologies, use ontology to capture expertise, insights, and understanding that is critical in facing complex issues and improve business processes. They use data that has been analyzed, interpreted, and contextualized (cleaned) to extract insights, patterns and actionable intelligence. This may include raw data, metadata, annotations and expert opinions.

In practice, knowledge systems lead to knowledge discovery, inference, reasoning and decision support. They incorporate technologies like knowledge graphs and artificial intelligence, (machine learning and natural language processing) to facilitate understanding and using information strategically.

Mining-tailored knowledge systems organize data and enrich it with context to transform it into information. New information, experiences and insights are continuously added to generate and capture knowledge. This enables the identification of new efficiencies that were not possible to identify by the human brain alone and provides real-time actionable information delivered to any decision maker, when they need it, across any domain of the mining process.

These systems can automatically absorb and contextualize data with the help of a domain ontology, and then use advanced analytics, AI, and Generative AI, such as machine learning and Large Language Models (LLM), to make critical decisions, such as potentially switching off a SAG mill based on sensor data processed through predictive failure algorithms. The Knowledge System acts as the central coordinator for all sub-systems controlling mine operations like mobile equipment, process plants, maintenance, and comminution circuits.

Superior productivity, efficiency, safety and ESG excellence are not attainable with just raw data. The mine of the future requires a knowledge system to process data in real time and to empower mines to act proactively, not retroactively.

About the author: Rudy Moctezuma is chief business relations officer of Eclipse Mining Systems.

Related posts