Open Collections

Machine learning on the correlation between weather condition and seepage rate : a case study on 20 years monitoring data from a waste rock dump

British Columbia Mine Reclamation Symposium; University of British Columbia. Norman B. Keevil Institute of Mining Engineering

Reliable prediction of seepage flow rate is important to flood controls and contaminant treatment for waste rock dumps. In this paper, a machine learning algorithm is developed to automatically study the correlations between seepage flow rate and mine site weather condition. Compared with traditional water balance approach, the advances of this study lie in all processes in the hydrological cycle require no simplification and assumption. Furthermore, a computer tool from the proposed approach is developed by Matlab and further applied to investigate a specific case on the full-scale waste rock dump of the Equity Silver mine. Seepage flow rates and weather conditions (precipitation and temperature) during 1998-2017 are used to train the tool accordingly. To validate this approach, the seepage flow rate predicted by the trained tool is compared with the real site monitoring data for that 20 years, which shows high agreement. It is also found that all full-scale peak flow scenarios in the field are fully captured and predicted by the developed computer tool. Overall, seepage flow rate was impacted mainly by long term weather conditions in the field, best represented in this case study by average weekly total precipitation and mean temperature of the preceding eight weeks.

Conference Paper

eng

Vancouver : University of British Columbia Library

10.14288/1.0391910

Non UBC

Other

DSpace