Abstract

Wet muck is a problem in underground mines due to the consequences it brings to the safety of workers, equipment, mining infrastructure, drawpoints, production drifts, and productive sectors, which can generate a loss in reserves. This paper describes a mathematical model to estimate wet muck entry for long-term planning applications at El Teniente. Three basins of El Teniente were included in the study of wet muck control: North, Centre, and South. Each basin includes mines with different characteristics in each exploitation sector. Consequently, models were built for each of the basins to represent its distinct reality. The models have been embedded in machine-learning software that estimates hazards associated with the extraction process for underground mines. To create the models, several variables – each associated with historical extraction – were investigated, including the extraction ratio and the height of draw, amount of water entering the cave, season of the year, presence of mud in neighbouring drawpoints, sectors closed due to wet muck above, and changes in surface or depressions. This study also includes granular flow variables and lithologies. In addition, fragmentation was included and estimated with a granular flow simulator, and the information was validated and calibrated with data from the El Teniente mine. Results indicate that our classification models can reproduce the wet muck phenomenon with an acceptable precision between 69% and 75% and an average tonnage error per drawpoint of 6%−15%. The approach detailed here to create models could