Mechanism characteristic analysis and soft measuring method review for ball mill load based on mechanical vibration and acoustic signals in the grinding process

• Published in: Minerals engineering 2018-11, Vol.128, p.294-311
• Main Authors: Tang, Jian, Qiao, Junfei, Liu, Zhuo, Zhou, Xiaojie, Yu, Gang, Zhao, Jianjun
• Format: Article
• Language: English
• Subjects: Frequency spectrum data; Mechanical vibration and acoustic signals; Mill load; Mineral grinding process; Soft measurement
• ISSN: 0892-6875; 1872-9444
• DOI: 10.1016/j.mineng.2018.09.006

•Mechanism analysis for operational expert cognitive and ball mill numerical simulation are reviewed.•The modeling process is summarized as a class of intelligent selective ensemble modeling problem.•The emphasis is the applied soft measuring strategies for mill load parameter measurement.•Possible directions for mill load soft measurement techniques are provided for future research. An operational optimization control for a mineral grinding process is limited by unmeasured load parameter inside a ball mill given its complex and unclear production mechanism. A mechanism characteristic analysis and soft measuring method for mill load parameter based on mechanical vibration and acoustic signals in the mineral grinding process is reviewed in this study. The modeling process based on the mechanical vibration and acoustic signals for the mill load parameters are summarized as a class of intelligent selective ensemble modeling problem. The applied soft measuring strategies for mill load parameter measurement are divided into three types, namely, off-line modeling, online modeling, and virtual sample generation, followed by a detailed discussion. Possible directions for mill load soft measurement techniques are provided for future research. These techniques include a vibration mechanism-based multi-component signal analysis, off-line intelligent ensemble soft measuring model based on simulation operational expert cognitive process, online updating strategy based on intelligently identified samples, and a mill load status intelligent recognition mode based on reinforcement learning strategy.