Optimizing mill performance through machine learning empowered sensing data analytics

点击翻译按钮获取中文摘要

Semi-autogenous (SAG) mills are employed in mining operations to break ore into a uniform aggregate size for further processing. The physical grinding process is carried out in a large drum where the ore and steel balls are added. To prevent damage to the drum, metal or liners are inserted. The costs of liner replacement depend on mill downtime and replacement parts so it is economical to change the liner as late as possible, but also maximize the mill output. Thus, the objective of this research is to maximize both the liner service life and the mill production. To achieve this goal, it is critical to predict the liner wear through the online monitoring during operation. This research will perform a comprehensive analysis of the data acquired by multiple monitoring sensors, such as vibration, strain, and bolt tension measurements. Particularly, we plan to use an electromagnetic acoustic transducer (EMAT) patented by the partner organization, Promet.ai, to measure the liner thickness in real-time. The EMAT sensor has the advantages to operate in field. The EMAT measurement together with other sensor data will be fused to predict the liner condition and remaining useful life, which is utilized to design a strategy for optimal control of the mill system. Data-driven modelling and deep learning based approaches will be explored in this study to achieve the predictive maintenance of the grinding mill in mining industry.