AI USE CASE
Ore Grade Prediction and Blending Optimization
Predict ore grade from drill data and optimize blending for consistent mill feed quality.
What it is
Machine learning models trained on drill hole assay data predict ore grade across the deposit, enabling optimized blending strategies that maintain consistent feed quality to the processing plant. Operations teams can reduce grade variability by 20–35%, cutting processing inefficiencies and reagent waste. Improved feed consistency typically lifts mill throughput by 5–15% and reduces unplanned downtime linked to grade swings. The result is more predictable recovery rates and lower cost per tonne processed.
Data you need
Historical drill hole assay data with spatial coordinates, ore type classifications, and processing plant feed quality records over at least 12–24 months.
Required systems
- erp
- data warehouse
Why it works
- Close collaboration between geologists, mine planners, and data scientists to validate model outputs against domain knowledge.
- Incremental deployment starting with one pit or ore zone to demonstrate value before scaling.
- Real-time or near-real-time data pipeline from drill rigs and assay labs into the prediction model.
- Change management program to build operator trust in AI-driven blending recommendations.
How this goes wrong
- Drill hole data is too sparse or inconsistently sampled, degrading model spatial accuracy.
- Models are built on historical geology that doesn't generalize to new mine faces or ore types.
- Blending recommendations are ignored by operators due to lack of trust in model outputs.
- Integration between the geological model and the plant scheduling system is missing, preventing operational use.
When NOT to do this
Don't deploy this when drill hole spacing exceeds the grade variability scale — predictions will be unreliable and blending targets unachievable.
Vendors to consider
Sources
This use case is part of a larger Data & AI catalog built from 50+ enterprise transformation programs. Take the free diagnostic to see how it ranks against your specific context.