AI Model Wiki

    Skip to main content
    18 min
    Lesson 2 of 6

    Ore Grade Prediction

    From the course AI for Mining & Resources

    Use AI to predict ore grades and optimize extraction value.

    Grade Estimation with ML

    class OreGradePredictor:
        def __init__(self):
            self.model = self.load_trained_model()
        
        def predict_grade(self, drill_data, location):
            features = {
                'geology': self.encode_geology(location),
                'drill_rate': drill_data.penetration_rate,
                'vibration': drill_data.vibration_signature,
                'torque': drill_data.torque_profile,
                'chip_color': self.analyze_chip_image(drill_data.camera),
                'neighbors': self.get_neighbor_grades(location, radius=50)
            }
            
            grade_prediction = self.model.predict(features)
            confidence = self.model.predict_confidence(features)
            
            return GradePrediction(grade_prediction, confidence)
    

    Real-Time Grade Control

    Sensor-Based Sorting

    1. Material passes through sensor array
    2. AI classifies grade in milliseconds
    3. Diverter routes to appropriate stockpile
    4. High-grade → Processing
    5. Low-grade → Waste

    Result: 10-20% increase in recovered value

    3D Ore Body Modeling

    def build_grade_model(drill_holes, assay_data):
        # Traditional kriging baseline
        kriging_model = OrdinaryKriging(drill_holes, assay_data)
        kriging_predictions = kriging_model.interpolate(grid)
        
        # ML enhancement with geology features
        ml_model = GradientBoosting()
        ml_model.fit(X=combine_features(drill_holes, geology_data), y=assay_data)
        ml_predictions = ml_model.predict(grid)
        
        # Ensemble combination
        return ensemble(kriging_predictions, ml_predictions, weights=[0.4, 0.6])
    

    Key Takeaways

    • •Grade prediction uses drill data, vibration, torque, and chip image analysis
    • •Real-time sensor-based sorting increases recovered value by 10-20%
    • •3D ore body models combine geostatistics (kriging) with ML for better accuracy
    • •AI grade control enables selective extraction of high-value material
    0 of 6 lessons marked complete

    Progress is kept in this browser only.