The recent projects from the IMS lab are categorized into four category of (1) production optimization, (2) predictive maintenance, (3) mine safety and (4) sustainability.
I. Production Optimization & Fleet Management
This research stream focuses on maximizing mine output, reducing operational costs, and enhancing efficiency through intelligent automation and optimized resource allocation.
- Intelligent Haul Truck Dispatch Under Dynamic Conditions: Some of the work can be found at https://doi.org/10.1016/j.resconrec.2022.106664 and https://doi.org/10.1016/j.jclepro.2023.140459. Moreover, we have filed a patent application for an autonomous dispatch system designed to optimize vehicle allocation in mining operations.
- Open Pit Production Scheduling: Long-term open pit mine planning works include extraction sequencing and block routing, integration of landfilling considerations into mine planning and mining-mineral processing integration using unsupervised machine learning. For short-term mine planning, a dig-limit optimization model was developed using mixed-integer programming.
- Stope Layout Optimization and Scheduling: A suite of approaches has been developed for stope layout optimization and scheduling using heuristic [1] [2], meta-heuristic [3] and linear programming [4] methodologies.
- Other Projects: Robust Mine Schedule Optimization
II. Equipment Health & Predictive Maintenance
This research area aims to minimize downtime, reduce maintenance costs, and improve the overall reliability of mining equipment through advanced predictive analytics.
- Ongoing Projects: Predicting Turbocharger Failures and Anomaly Detection
III. Mine Safety
This stream investigates the use of AI and data analytics to improve mine safety, enhance hazard identification, and mitigate operational risks.
- Cave Mine Pillar Stability Analysis using Machine Learning: This study is an ongoing collaboration with another professor in the department, Prof. Steve McKinnon. Current pillar stability assessments in large-scale cave mines primarily rely on stress analysis, which often struggles to account for the complex influence of operational factors. This project, conducted as a case study at the Chuquicamata underground cave mine, applies a tree-based machine learning approach to evaluate the impact of these operational conditions on pillar stability. We have achieved 80% accuracy in predicting pillar collapses despite limited data availability, which was a result unattainable through traditional stress analysis methods.
- Development of Targeted Safety Hazard Management Plans: This project uses multidimensional association rule mining to identify relationships between operational factors and safety hazards, enabling the creation of targeted safety plans to prevent incidents and improve worker safety.
- Short-Term Rockburst Risk Prediction: Research in this area focuses on developing ensemble learning methods for accurate short-term rockburst prediction. Utilizing microseismic data from the Jinping II hydropower station, the models achieved up to 86.67% accuracy in predicting rockburst risk, outperforming individual base learners.
- Cognitive Work Analysis Through Latent Variable Modeling: Investigating cognitive workload and decision-making processes in mining operations, this research aims to improve safety protocols and reduce human error. By critically comparing exploratory, confirmatory factor analysis, and structural equation modeling, the study highlights the potential to identify key factors impacting mine safety and improve accident research.
IV. Sustainability & Decarbonization
This stream is dedicated to developing technologies and strategies to reduce the environmental footprint of mining operations. These projects aim to reduce greenhouse gas (GHG) emissions and improve overall environmental sustainability.
- Environmental and Economic Analysis of Electric Vehicle Transition: Collaborating with Prof. Qian Zhang, we investigated the potential of transitioning to all-electric haulage fleets to reduce GHG emissions in the mining sector and achieve net-zero objectives. Building a discrete event simulation model, we compared diesel and electricity demands, demonstrating potential GHG emission reductions of up to 92.6% and operating cost savings of 40-62% with electric trucks.
- Other Projects: Machine Learning Mineral Process Optimization & Decarbonization and Reducing Fuel Consumption in Haul Trucks