ToolFusion
πŸ“‹ Case Study

Copper Mine Block Model Refinement Using Neural Kriging

Traditional kriging over-smoothed high-grade chalcocite zones, causing 8.2% reserve underestimation

πŸ—οΈ Project Overview

Escondida-style porphyry copper deposit, Chile

🎯 Challenge

Traditional kriging over-smoothed high-grade chalcocite zones, causing 8.2% reserve underestimation

πŸ”§ Design Approach

Hybrid neural network trained on 3D variogram features + geochemical pathfinder ratios; integrated with Surpac via Python API

πŸ“ Design Diagram

Copper Mine Block Model Refinement Using Neural Kriging Traditional kriging over-smoothed high-grade chalcocite zones βˆ’8.2% reserve Neural Kriging Engine 3D variogram features + geochemical pathfinder ratios Surpac Integration Python API β€’ Real-time update RMSE Reduction 1.7 β†’ 0.9 g/t Reserve Upside +12.4 Mt @ +0.18% Cu

AI-generated project design illustration

πŸ“ Key Calculations

RMSE Reduction

RMSEβ‚œα΅£β‚π’Ή βˆ’ RMSEₙₑᡀᡣₐₗ
Result: 1.7 g/t β†’ 0.9 g/t
Improved high-grade continuity capture

Reserve Upside

Ξ”Tonnage Γ— Ξ”Grade
Result: +12.4 Mt @ +0.18% Cu
Extended mine life by 2.3 years

πŸ“Š Results

12.4 Mt reserve addition, 92% reduction in false-negative ore classification, 18-month payback on AI infrastructure

πŸ’‘ Lessons Learned

  • β€’Domain-specific feature engineering outperformed off-the-shelf CNNs
  • β€’Human-in-the-loop validation preserved geological plausibility
  • β€’GPU-accelerated inference enabled daily block model updates

βœ… Key Takeaways

  • 1Domain-specific feature engineering outperformed off-the-shelf CNNs
  • 2Human-in-the-loop validation preserved geological plausibility
  • 3GPU-accelerated inference enabled daily block model updates