Improving Resource Estimation and Reconciliation with Machine Learning

Models quantifying the grade and tonnage of mineral deposits form the basis of important and costly decisions for planning, optimization and extraction of a natural resource. Models are initially generated from sparse exploration sampling; however, information is continuously collected until resource extraction. Predicted values that reconcile well with true values following extraction instill confidence in the production forecasts. Failure to meet production forecasts can have crippling effects on cash flow and ultimately result in failure of the project.
In this research a neural-network-based prediction framework is proposed that incorporates production information to the predictive algorithm to improve forecasts of future production, thereby improving reconciliation at a mining project. The proposed method could be used to continually update resource models to improve decisions being made at all scales. This research will benefit the partner company since the incorporation of a wide array of data in manual reconciliation is complex. The proposed research will simultaneously simplify the workflow for the practitioner and improve reconciliation by improving predicted values in unmined areas. This will generate value through increased operational efficiency.

Faculty Supervisor:

Jeff Boisvert


Ryan Martin


Teck Resources Ltd


Engineering - petrochemical


Mining and quarrying




Current openings

Find the perfect opportunity to put your academic skills and knowledge into practice!

Find Projects