Highlights from COSMO Technical Day 2025
On June 20th, our team attended the 2025 COSMO Technical Day—an event that brought together top researchers and industry leaders advancing the frontiers of stochastic mine planning, geostatistics, and AI in mining.
The day offered a rich mix of cutting-edge research and real-world applications, showcasing the latest advancements in stochastic mine planning, advanced geostatistical simulation techniques, and the integration of AI into stochastic planning. It underscored the growing importance of moving beyond deterministic assumptions and adopting stochastic methods to more effectively quantify and manage geological and operational uncertainties. The event highlighted both the tangible impact of these innovations on mining operations and the cutting-edge research emerging from the COSMO Lab at McGill University.
Presentations of Industry Applications
Some real-world case studies from mining companies showed how stochastic methods and geostatistical simulations are already being used to unlock value, manage uncertainty, and improve decision-making:
KPI Mining Solutions
Stochastic Optimization of a Low-Grade Iron Ore Mining Complex
A case study led by the KPI Mining team focused on a low-grade deposit, where the conventional deterministic plan revealed significant risks, particularly related to sulfur contamination and the forecasted iron head grade. Additionally, the potential for dilution and ore loss was identified. Using the stochastic approach, the team developed a more robust mine plan, from which risk-resilient operational phases were designed. Ultimately, re-optimization within phases was conducted (again using the stochastic method) to ensure maximum value. The final designed mine plan showed 25% higher NPV compared to the conventional base case.
Vale Base Metals
Reconciling Models and Plans with Reality
A real-world case study was presented reconciling both the stochastic and deterministic mine plans against a mined-out area with available blasthole data (as the best proxy for reality). Results showed the greater adherence of geostatistical simulations with reality, as well as the higher predictability of grade and tonnage forecasts that came from implementing the stochastic optimization using simulations. Infill Drilling Optimization
One case study highlighted the power of stochastic approaches for infill drilling optimization and pattern definition.
Seismic Risk in Underground Mining
Stochastic methods were also applied to manage uncertainty around seismic activity in deep underground mines, enabling more reliable and less interrupted production plans with increased value. Vale S.A.
Managing Deleterious Elements with Stochastic Methods
In one case study presented on an iron ore mining complex, the risk analysis of the deterministic plan identified a high probability of underestimating contaminant grades. To mitigate such risks, stochastic plans were developed, leading to improved product specification control. The stochastic results also showcased the potential for higher Fe concentrate production, higher NPV, and longer LoM. Ongoing Research at the COSMO Lab
Integrating Long- and Short-Term Planning Under Uncertainty
One outstanding project focused on applying stochastic methods to long- and short-term planning in an integrated model, considering uncertainties in plant performance and geology, while incorporating CAPEX decisions. This model ensures operations can hit their targets across multiple time horizons. Reclamation Integration & Sustainability
One of the most fascinating topics was integrating progressive reclamation directly into the stochastic long-term planning process. A framework was developed to manage potentially acid-generating waste material using encapsulation with non-acid-generating waste material. In one copper mine, this reduced haulage cost per ton mined by 15% and increased NPV by 4%, showing that sustainability can enhance profitability. End-to-End Learning for Mining Optimization
Another AI-focused research worked on an end-to-end learning framework for mining optimization which can be trained on specific projects to make future optimization runs more effective and efficient. Short-Term Planning with Geological Uncertainty
Another exciting project optimized fleet management systems given geological uncertainties at the short-term planning scale. Sustainable Fleet and Short-Term Optimization
Another stochastic optimization framework jointly optimized short-term planning with fleet management with the aim of reducing carbon footprint while maintaining profitability, showcasing that operational efficiency and sustainability can go hand in hand. Advancements in Geostatistical Simulations
The ongoing research did not overlook the advancement of geostatistical simulation techniques. One study was on the advancements of high-order simulation methods, reproducing high-order statistics of the data, and avoiding the Gaussian assumptions. Another used a generative model based on pattern recognition, eliminating the need for an explicit density function. Underground Mining Optimization
Underground mining was also another theme. Advanced and efficient solution methods are being developed for joint stochastic optimization of stope design and scheduling, and are being tested with promising results.
At KPI Mining, we’re excited to see the industry advancing toward more robust, uncertainty-aware mine planning—an evolution that directly supports our mission. We’re proud to help mining companies generate higher-value, risk-resilient plans by making advanced, research-backed technologies not only accessible, but also practical and impactful across a wide range of operations. By translating innovation into proven solutions, we enable our partners to unlock value, mitigate risk, and plan with greater confidence.