Abstract
Reliable prediction of drainage flow and chemistry from waste rock storages is important to the flood control and contaminant treatment at many mine sites. At this time, the state-of-art predictive models generally require identifications of the physical processes inside waste rock storages, and also intensive characterization. However, the prediction accuracy is not satisfied sometimes as many simplification and assumptions are adopted in the prediction. In the big data era, various sensors have been put in the field to monitor mine sites, which enable advanced machine learning technology to investigate those monitoring data and study the underlying patterns behind them. This presentation will introduce how National Research Council Canada use artificial neural network to predict the drainage flow and chemistry from waste rock storages based on historical monitoring data. The advantage of this technology is that no further costly characterization are required to make a reliable prediction, as most geo-bio-chemical mechanisms involved in waste rock storages have been naturally implied in the monitoring data and they can be captured through machine learning process.