The Future of Orebody Knowledge
What we know is important. What we don't know is often more important. Mining has long focused on quantifying uncertainty in the orebody. The next challenge is understanding uncertainty in our knowledge of the orebody itself.
This page explores emerging ideas, research directions and future capabilities aimed at improving decision confidence across geology and geometallurgy.
Data is not knowledge.
Mining operations continue to collect increasing volumes of geological, geometallurgical and operational data. Yet critical decisions are often constrained not by the amount of data available, but by gaps in understanding, incomplete context and uncertain assumptions. Future analytical systems must do more than produce predictions. They must help identify what is known, what is not known and where further evidence is required.
Mining companies spend millions reducing geological uncertainty through drilling. But they spend comparatively little quantifying knowledge uncertainty.
Imagine a technical manager hearing:
"We have 10 million assay records."
That sounds reassuring.
Compare with:
We can confidently answer 78% of critical orebody questions.
14% are partially supported.
8% have insufficient evidence.
That is a completely different conversation.
This second statement is actionable.
From prediction to decision confidence.
Many analytical systems answer: What is the estimate?
Future systems should also answer: How much should we trust the estimate?
Understanding the level of evidence supporting a prediction may be as important as the prediction itself.
- Areas of interest include:
- evidence assessment
- uncertainty communication
- confidence-informed decision support
- knowledge gap identification
Measuring orebody knowledge.
Geological understanding evolves over time through drilling, sampling, interpretation, reconciliation and experience. An important future challenge is determining how well that knowledge is captured, retained and applied.
- Areas of exploration include:
- orebody knowledge assessment
- knowledge coverage analysis
- knowledge gap mapping
- organisational knowledge retention
- decision support frameworks
Knowledge confidence.
Resource confidence is widely measured across the mining industry.
Knowledge confidence is rarely measured.
The objective is not to eliminate uncertainty.
The objective is to better understand it.
- Future approaches may seek to evaluate:
- whether critical questions can be answered
- how strongly those answers are supported
- where assumptions dominate evidence
- where additional data would most effectively reduce uncertainty
Current areas of investigation.
Assessing how strongly available evidence supports technical conclusions.
Identifying unanswered questions and areas requiring additional investigation.
Preserving, organising and operationalising geological knowledge across the life of an asset.
Developing approaches that operate within strict data governance and security requirements.
"What if every study, interpretation, model update, reconciliation exercise and review performed over the last decade remained available through a site-specific expert system that could explain reasoning, identify uncertainty and preserve institutional knowledge?"
Looking ahead.
The mining industry has become increasingly effective at collecting and storing data. The next opportunity lies in improving how knowledge is measured, retained and applied. Future systems should support not only prediction and analysis, but also understanding, confidence and informed decision making.
1. Prediction - What is the answer?
2. Explanation - How was this answer generated?
3. Trust - How much should I believe this answer?
Can the internal representation (IR) be used to build a knowledge-mining system?
Yes.
The internal representation IR could be used to train two specialised language models (SLMs):
- A feasibility model that determines whether a question can be answered.
- A generative model that learns the engineered relationships and produces the answer.
Because the IR is encoded, the SLM can run in the cloud and be wrapped in an API
without exposing operational data.
Example prediction with decision support.
Prediction:
Recovery = 91%
Can the site's knowledge base explain this result?
Yes
Supporting evidence:
- Ore mineralogy
- Liberation characteristics
- Historical testwork
- Similar ore domains
Knowledge Coverage: 87%
Prediction:
Recovery = 91%
Can the site's knowledge base explain this result?
No
Missing evidence:
- Recent variability testing
- Domain-specific metallurgical response
- Fresh ore transition behaviour
Example prediction with reasoning and uncertainty.
Estimate: 1.5 g/t Au
Trust Level: High
Reasoning:
- Within trained data range
- Similar geological contexts exist
- Supported by adjacent drilling
- Consistent with domain trends
Primary uncertainty:
- Local fault interpretation
Estimate: 1.5 g/t Au
Trust Level: Low
Reasoning:
- Significant extrapolation
- Limited nearby drilling
- Geological analogues weak
- Similar cases not present in training data
Recommendation:
- Additional drilling required
Visit the mining-ds ecosystem for a collection of useful tools, source code, example analysis, and research.
Interested in discussing these ideas?
Whether your challenge involves geology, geometallurgy, data quality or decision support, let's talk.
Request a Chat