Applying machine learning
If you’ve been following this series of articles, you will know that a number of machine learning pilot projects have already demonstrated that AI techniques based on machine learning can help the mining industry make better, more informed decisions. In my view, that means the question now becomes not whether to adopt it, but how to integrate it into our regular processes.
The use-cases and success stories that follow illustrate how the mining industry can employ machine learning to take full advantage of massive amounts of (often underused) mining data to, among other things:
- maximise ore recovery
- evaluate material better
- improve production
- reduce variability, and
- increase revenue.
Use-case #1: Classifying and predicting mineral flotation performance
We know that flotation performance has a direct impact on ore recovery but predicting it can be challenging.
Machine learning allows us to integrate the geological variables of the incoming material with the real-time operational variables to create a single workflow for data cleaning and pre-processing. From this single workflow, we can then select the best variables to use in algorithms in order to identify and classify patterns in the data not detectable by humans alone.
With those patterns now classified, we can:
- create multiple scenarios of mineralogy and operating variables
- cluster them, and
- perform a recovery analysis and definition of thresholds for each cluster.
Finally, we can do predictive analysis to find the one scenario where the probabilities most exceed the success threshold, and from there determine the report parameters and behaviors capable of exceeding the success threshold.
Use-case #2: Improving rock characterisation
Incorrect rock characterisation can lead to material destination errors, including ore underestimation or material-to-waste-dump overestimation, which in turn can lead to potentially negative effects on the environment. However, because material tracking is a multi-variable process and we often lack detailed geological logging data, it’s easy to make mistakes.
Machine learning can help.
By training machine-learning algorithms based on mineralogical data, we can characterise rocks even with limited information on the deposit and, for each block of the solids model, estimate the probability of the rock containing elements that can harm the environment. This will not only make our treatment decisions more accurate and efficient, it will also ensure that sustainability is built-in to the project from the beginning.
Success story #1: CODELCO
CODELCO, Chile’s largest state-owned mining company, operates El Teniente, the world’s biggest underground copper mine.
In recent years, the company has increased El Teniente’s SAG mill ore processing 2% annually by using machine learning to analyse big data and create digital models. That may not sound like a lot, but a 2% annual increase means somewhere around US\$30M in additional revenue per year.
Success story #2: BHP
In operation since 1885, BHP is one of the biggest mining companies in the world. Using machine-learning models that predict grade biases daily and update mine plans and targets accordingly, it has reduced iron ore grade variability across its Australian Jimblebar iron ore operation, resulting in a US\$10 million increase in revenue.
BHP is also using machine learning algorithms to analyse component failure history and wear on engine components in real-time as part of its Maintenance Centre of Excellence. This Center is dedicated to evaluating the entirety of the company’s machinery data, across all operations, to predict equipment maintenance needs.
In other words …
Machine learning can add significantly to your knowledge base and lead to better-informed decision-making that results in increased revenue.
In my final article, I will provide useful tips for implementing your own machine learning projects.
About the Author
Jose Gonzalez is a Geologist from the University of Chile. Experience in modeling, participation in projects of economic geology, geostatistics and sustainability in mining. Applied knowledge in Data Science, oriented to the innovation of digital solutions. José González has conducted trainings in the LATAM region, supporting numerous projects and his role is oriented to the search and adoption of new solutions that can strengthen the mining services portfolio.
Related posts:
