Use of time series event classification to control ball mill performance in the comminution circuit - a conceptual framework

Metallurgical attributes are often omitted from the mine to metal valuation models since they are either absent or unreliable. However, recent developments in sensor technology indicate the potential to collect information on metallurgical properties directly or by measurement of proxies. Integrating this information back into the resource model would provide the necessary means to move towards a more comprehensive and reliable evaluation model. To obtain truly optimized mining decisions it is necessary to consider the metallurgical attributes since they are indicated as root cause of changing plant performance. Therefore, a better metallurgical characterization of the plant feed over time is required, which allows for a more optimal selection of process control settings. Different material types have varying effects on machine performance in the comminution circuit. This makes it possible to refer a performance change as a response to different geological attributes. Hence, the corresponding geological machine behaviour can be controlled by defining effects of behavioural geology. This paper introduces a framework containing data fusion of sensor responses which resemble geological attributes and subsequent multivariate time series machine behaviour characterization for improved process control in the comminution circuit. The conceptual framework’s approach is that process control in future will be supervised by profound knowledge from sensor data indicating geological behaviour. The use of multivariate time series deep learning is proposed to create innovative process control. This innovative control is then a response to a combination of advanced sensor data (XRF, LIBS, FTIR, etc.) with more traditional sensor data (throughput, density, etc.). These advanced sensors provide more knowledge about material specific properties in the form of discoverable events. This new knowledge is important in the vision of behavioural geology, to better understand the influence of geological behaviour on machine performance.